<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Zach Edwards on Medium]]></title>
        <description><![CDATA[Stories by Zach Edwards on Medium]]></description>
        <link>https://medium.com/@thezedwards?source=rss-e55d568b3a27------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*Vhci1IFb9SXsDeIa1B5atg.jpeg</url>
            <title>Stories by Zach Edwards on Medium</title>
            <link>https://medium.com/@thezedwards?source=rss-e55d568b3a27------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 15:27:45 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@thezedwards/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Breitbart.com is Partnering with RT.com & Other Sites via Mislabeled Advertising Inventory]]></title>
            <link>https://medium.com/@thezedwards/breitbart-com-is-partnering-with-rt-com-other-sites-via-mislabeled-advertising-inventory-6e7e3b5c3318?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/6e7e3b5c3318</guid>
            <category><![CDATA[ad-fraud]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[ads-txt]]></category>
            <category><![CDATA[advertising]]></category>
            <category><![CDATA[marketing-strategies]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Wed, 22 Jul 2020 19:37:47 GMT</pubDate>
            <atom:updated>2020-07-23T01:36:25.429Z</atom:updated>
            <content:encoded><![CDATA[<h4>A large group of alt-right sites, low quality publishers, and other websites are mislabeling Ads.Txt publisher relationships and potentially committing a form of advertising fraud.</h4><p><strong>Summary: </strong>The Interactive Advertising Bureau’s ads.txt standard is being abused by publishers mislabeling and sharing a “DIRECT” label for account-bidding IDs used in online bidding protocols — with the DIRECT labels being spread out across sometimes hundreds of unrelated websites. This inventory mislabeling creates online advertising Dark Pool Sales Houses, a market behavior that likely violates consumer protection frameworks, and in the future, securities laws should be expanded to explicitly address the purposeful mislabeling of online advertising inventory.</p><h4>TABLE OF CONTENTS</h4><ul><li>Introduction to Overlapping DIRECT IDs in an Ads.txt Record — Breitbart.com &amp; RT.com</li><li>The origins of ads.txt</li><li>Ads.txt was supposed to reduce online advertising fraud</li><li>Ads.txt uses certification bodies to check sales houses — but they don’t work</li><li>Ads.txt was adopted slowly, until Google started pushing it</li><li>Breitbart is aggressively mislabeling DIRECT inventory, pooling money and data into Dark Pool Sales Houses</li><li>Advertisers can’t tell if they’ve been funding Breitbart</li><li>Breitbart doesn’t label their Dark Pool Sales Houses — but RT.com does!</li><li>Many companies are helping publishers create Dark Pool Sales Houses…or looking the other way as it happens…</li><li>Here are 34 websites that share DIRECT ID’s with Breitbart</li><li>Dark Pool Sales Houses need more scrutiny from the advertising industry and regulatory oversight</li><li>Conclusion &amp; Next Steps</li><li>Online Advertising Brand Problems Require Businesses to “Check Your Ads” — and Why Co-Founders Nandini Jammi and Claire Atkin Understand Brand Risk in the Online Ads Space in the Era of Fake News, Disinformation, and Hate Speech</li></ul><p>This research is being published in collaboration with Nandini Jammi and Claire Atkin, co-founders of Check My Ads (<a href="https://www.checkmyads.org/">https://www.checkmyads.org/</a>) and publishers of <a href="https://branded.substack.com/">the must-subscribe newsletter BRANDED</a>. <a href="https://branded.substack.com/p/so-thats-how-breitbart-is-still-making">Click here to view their summary and details</a>.</p><h3>Introduction to Overlapping DIRECT IDs in an Ads.txt Record — Breitbart.com &amp; RT.com</h3><p>If you compare the<a href="https://www.breitbart.com/ads.txt"> Breitbart.com/ads.txt</a> file against the<a href="https://www.rt.com/ads.txt"> RT.com/ads.txt</a> file, you will find at least 4 identical account-bidding IDs that share the “DIRECT” label on both sites, which is mislabeled RESELLER inventory being sold as DIRECT inventory to advertising buyers.</p><p>These account-bidding IDs are also known as “SellerIDs” “AccountIDs” or “BiddingIDs” — several industry terms relate to this same ID found in the ads.txt file.</p><p>Any shared DIRECT account-bidding IDs across uniquely owned websites creates a Dark Pool Sales House that spreads money and data across all the websites participating in the architecture.</p><p>The account-bidding IDs found in the ads.txt records for RT.com and Breitbart.com that both say “DIRECT” and are otherwise identical, are:</p><blockquote>pubmatic.com, 154037, DIRECT, 5d62403b186f2ace (<a href="https://www.google.com/search?q=%22pubmatic.com%2C+154037%2C+DIRECT%2C+5d62403b186f2ace%22+-site%3Awww.adstxtvalidator.com"><strong>Dark Pool Sales House across ~208 websites</strong></a>)</blockquote><blockquote>appnexus.com, 8233, DIRECT, f5ab79cb980f11d1 (<a href="https://www.google.com/search?q=%22appnexus.com,+8233,+DIRECT,+f5ab79cb980f11d1%22+-site:www.adstxtvalidator.com"><strong>Dark Pool Sales House across ~101 websites</strong></a>)</blockquote><blockquote>appnexus.com, 9393, DIRECT, f5ab79cb980f11d1 (<a href="https://www.google.com/search?q=%22appnexus.com,+9393,+DIRECT,+f5ab79cb980f11d1%22+-site:www.adstxtvalidator.com"><strong>Dark Pool Sales House across ~194 websites</strong></a>)</blockquote><blockquote>indexexchange.com, 186684, DIRECT, 50b1c356f2c5c8fc (<a href="https://www.google.com/search?q=%22indexexchange.com,+186684,+DIRECT,+50b1c356f2c5c8fc%22+-site:www.adstxtvalidator.com"><strong>Dark Pool Sales House across ~10 unique websites</strong></a>) — This is an extremely small Dark Pool Sales House, mostly just RT.com and Breitbart.com — the other sites are small)</blockquote><p><a href="https://docs.google.com/spreadsheets/d/e/2PACX-1vQXg6sJVh-L3r8OljGNLqLqklxqPnaJ38bog7U_9aAZs_h8PMUbIFCnk4Rg-tg5OtgZn5Wx8P6l7Lhk/pubhtml?widget=true&amp;amp;headers=false"><strong>Click here to view the Breitbart Ads.txt records as of July 2020 — View the most up-to-date version of this file via Breitbart.com/ads.txt …</strong></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V44OtXYUEJK1OpwCGN_raw.jpeg" /><figcaption><a href="https://iabtechlab.com/wp-content/uploads/2019/03/IAB-OpenRTB-Ads.txt-Public-Spec-1.0.2.pdf">The final IAB ads.txt standards</a> clearly state that a value of “DIRECT” indicates “That the Publisher (content owner) directly controls the account indicated in field #2 on the system in field #1.” The “RESELLER” field is also clearly defined as indicating “that the Publisher has authorized another entity to control the account indicated in field #2 and resell their ad space via the system in field #1.”</figcaption></figure><h3>The origins of ads.txt</h3><p>The IAB Tech Lab has established the specifications for ads.txt.</p><p>Their<a href="https://iabtechlab.com/wp-content/uploads/2019/03/IAB-OpenRTB-Ads.txt-Public-Spec-1.0.2.pdf"> ads.txt standards</a> state that a value of “DIRECT” indicates “that the Publisher (content owner) directly controls the account indicated in field #2 on the system in field #1.”</p><p>The “RESELLER” field is also clearly defined as indicating “that the Publisher has authorized another entity to control the account indicated in field #2 and resell their ad space via the system in field #1.”</p><p>While these standards are only 3 years old, currently in 2020, these standards are being abused, with the DIRECT label being applied to RESELLER inventories across numerous unscrupulous websites, a potentially fraudulent practice.</p><p><strong>Breitbart.com is among numerous websites abusing the global ads.txt standards created by the Interactive Advertising Bureau (IAB)</strong>, and due to their efforts to clone account-bidding IDs across domains, while mislabeling the “DIRECT” relationship field that is supposed to differentiate between a “RESELLER” inventory and the directly-controlled “DIRECT” inventory, online advertising buyers who have purchased ads on hundreds, if not thousands of websites have been quietly allowing their data and money to pool into Dark Pool Sales Houses that fund the fake news, disinformation and hate speech that is published on Breitbart News and similar publications, including RT.com.</p><p>Many of the websites deploying this Direct-Inventory-Mislabeling will be familiar names in conservative circles, including: Breitbart, Drudge Report, the Daily Caller, PJ Media, Fox9, and more, but many other organizations also have overlapping Direct-mislabeled inventory with Breitbart, including Vimeo.com, MSN.com, Upworthy.com, TalkingPointsMemo.com, NBA.com, Adweek.com, and other publishers and organizations that may not have realized that they are working with Breitbart to build a global online advertising mislabeling ring.</p><p>This process to create Dark Pool Sales Houses has<a href="https://digiday.com/media/ssp-arbitrage/"> historically been done through opaque SSP arbitraging</a>, which many buyers hoped would end with the deployment of ads.txt. From Digiday in 2017, before ads.txt:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mbwduY68OISyFb_w8HByrw.jpeg" /><figcaption>From Digiday in February 2017, “‘<a href="https://digiday.com/media/ssp-arbitrage/">We go straight to the publisher’: Buyers beware of SSPs arbitraging inventory</a>”</figcaption></figure><h3>Ads.txt was supposed to reduce online advertising fraud</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*q7T1D6x6yaZFXm3YwnIF9g.jpeg" /><figcaption>Version 1.01 of the Ads.txt Specification was <a href="https://iabtechlab.com/wp-content/uploads/2017/09/IABOpenRTB_Ads.txt_Public_Spec_V1-0-1.pdf">released by the IAB in September 2017</a></figcaption></figure><p>Ads.txt is a relatively new online advertising standard created in the Spring of 2017 by the Interactive Advertising Bureau (IAB), with<a href="https://iabtechlab.com/wp-content/uploads/2017/09/IABOpenRTB_Ads.txt_Public_Spec_V1-0-1.pdf"> the first versions of the ads.txt standard being released in November 2017</a> and then eventually<a href="https://iabtechlab.com/wp-content/uploads/2019/03/IAB-OpenRTB-Ads.txt-Public-Spec-1.0.2.pdf"> finalized in the Spring of 2019</a>.</p><p>The authors and contributors to the ads.txt standard are mostly well-known technologists leading large companies working in the online advertising space. A huge portion of the organizations who worked on the original standards have been letting their own clients abuse the bid labeling requirements, which is creating whole new layers of unscrupulous and deceptive advertising practices, and CPM clearing houses that couldn’t and didn’t exist before the ads.txt standards were deployed.</p><p>Ads.txt was meant to “clean up the online advertising industry.” It has done the opposite. The vast majority of advertising companies who have deployed any types of domain block lists for their clients have not also deployed tools to effectively find these new types of ads.txt Dark Pool Sales Houses and block those mislabeled DIRECT account-bidding IDs from their inventories.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nYOu6fobx9d6reyvsO3gtA.jpeg" /><figcaption>View the Ads.txt &amp; App.Ads.txt “standards” from the IAB @ <a href="https://iabtechlab.com/ads-txt/">https://iabtechlab.com/ads-txt/</a></figcaption></figure><p>Currently, ads.txt has been endorsed and deployed by all the major online advertising exchanges.<a href="https://wiki.iabtechlab.com/index.php?title=Ads.txt_Normalization_Mappings"> The IAB has a helpful list that includes dozens of the largest advertising exchanges</a> and their ads.txt mappings, but this list does not reflect even a tiny portion of the actual “advertising exchanges” (many of which are Dark Pool Sales Houses) that show up in publisher ads.txt files across the internet.</p><p>Prior to this standard, account-bidding IDs needed to be parsed through header bidding files, post-auction pixel/cookie syncs, and through other server-side methods working directly with ad exchanges. But now, anyone participating in online advertising on a domain needs to deploy a file located at domain.com/ads.txt with a consistent format, which is ingested and parsed to help online advertising buyers target their ads effectively, and help them avoid specific types of Reseller inventories that are typically lower quality than Direct inventory.</p><p>It’s always been common knowledge that advertising Dark Pool Sales Houses exist. The IAB mentioned that sales houses were used in Japan during ads.txt publisher discussions. That’s why the “RESELLER” label was created. <a href="https://community.pubmatic.com/display/QC/Ads.txt+Implementation+Guide">Pubmatic, too, includes a mention of Sales Houses on their ads.txt implementation page</a>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ByILbgdrkox3v406KUUsTg.jpeg" /><figcaption><a href="https://community.pubmatic.com/display/QC/Ads.txt+Implementation+Guide">Pubmatic includes a mention of Sales Houses on their ads.txt implementation page</a></figcaption></figure><h3>Ads.txt uses certification bodies to check sales houses — but they don’t seem to be working</h3><p>In every ads.txt record, there is a note mentioning a “certification authority.” This is typically (always?) from the “Trustworthy Accountability Group” (<a href="https://www.tagtoday.net/">TAG</a>) who were amongst the original authors of the ads.txt standards. Unfortunately, a huge portion of the mislabeled “DIRECT” account-bidding IDs found on Breitbart.com also have TAG-IDs, which is an indication that TAG has no transparency into Dark Pool Sales Houses, or isn’t banning them.</p><p>A TAG certification apparently doesn’t prevent their clients from deploying these mislabeled DIRECT IDs across hundreds of unrelated domains.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HCYjQOPlr_kwjBHfHys15g.jpeg" /><figcaption><a href="https://wiki.xandr.com/display/industry/Xandr+Support+for+Ads.txt+and+App-Ads.txt">Xander.com provides ads.txt information</a> for buyers and publishers, also provides some details on the ads.txt Certification ID oftentimes provided by TAG.</figcaption></figure><h4>Here’s the “Google Ad Manager Training” video titled “DoubleClick: Create an ads.txt File.”</h4><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FT_4uK2HnOy0%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DT_4uK2HnOy0&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FT_4uK2HnOy0%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/0cfe9043409004e22be5881e54773d5a/href">https://medium.com/media/0cfe9043409004e22be5881e54773d5a/href</a></iframe><p>In this video, <strong>Google specifically mentions the DIRECT label being used to indicate that the publisher directly controls the account</strong> , with their narrator saying:</p><blockquote>“…the third required field is the type of account or relationship. Use the value ‘DIRECT’ to indicate that the publisher directly controls the account indicated in field #2 on the system in field #1. This tends to mean a direct business contract between the publisher and the advertising system. Use the value ‘RESELLER’ to indicate that the publisher has authorized another entity to control the account indicated in field #2 and resell their ad space via the system in field #1.”</blockquote><p>Google has always advised clients to properly identify DIRECT vs RESELLER inventory, but it’s unclear if they are stopping or policing this practice.</p><p>It’s also unclear if Google’s “Bid time invalid traffic protection” is blocking bids from account-bidding IDs labeled as “DIRECT” when they are used in Dark Pool Sales Houses, but most public indications point to this <strong><em>not</em></strong> currently happening.</p><p><strong>From Google’s </strong><a href="https://www.blog.google/products/marketingplatform/360/working-with-industry-towards-fraud_21"><strong>ads.txt blog post</strong></a><strong>:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EKTuSoysI-tP2UgUao1xHQ.jpeg" /><figcaption>From Google’s “<a href="https://www.blog.google/products/marketingplatform/360/working-with-industry-towards-fraud_21">Working with the industry towards a fraud-free media supply chain</a>” blog post.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SrHCqWREoHXYMhCavJllIA.jpeg" /><figcaption>From Digiday @ “‘<a href="https://digiday.com/media/can-good-big-stick-thanks-google-ads-txt-taking-off/">It can be good to have a big stick’: Thanks to Google, ads.txt is taking off</a>”</figcaption></figure><h3>Ads.txt was adopted slowly, until Google started pushing it</h3><p>In 2017, adoption was slow until Google started to aggressively endorse the ads.txt protocols and notifying publishers they would restrict bids for websites that didn’t use the new standards.</p><p><a href="https://digiday.com/media/can-good-big-stick-thanks-google-ads-txt-taking-off/">Digiday wrote</a> “Before Google announced it would use ads.txt files to filter unauthorized inventory, just 20 percent of the publishers that work with ad tech firm Advelvet had ads.txt files, said Murat Deligoz, CEO of Advelvet, which helps publishers set price floors in Google’s ad exchange. Now, all the publishers that work directly with Deligoz’s company have ads.txt files. ‘<strong>Google is definitely the force to get ads.txt up and running</strong>,’ he said.”</p><p>In 2017,<a href="http://www.adopsinsider.com/ad-exchanges/ads-txt-adoption/"> AdOpsInsider’s Ben Kneen highlighted why the DIRECT label was going to be preferred for advertising buyers</a>, writing, “Buyers will most likely prefer direct supply paths to reseller paths, as there is one less company in the value chain and fees are therefore likely lower through direct paths.”</p><p>As ads.txt was deployed, in 2018<a href="https://digiday.com/marketing/state-ads-txt/"> articles started to be written about how the deployment of ads.txt was increasing CPM rates for premium publishers</a>. Few articles after this have parsed the ways ads.txt could be abused, or why it’s crucial for publishers to not share DIRECT labels with other businesses, without crossing an important line into mislabeling their own inventory.</p><p>The “RESELLER” label was created for a reason, but without anyone paying attention, the global advertising supply chain now has a mislabeled supply of DIRECT account-bidding IDs that online advertising buyers are being forced to navigate without the tools to avoid this inventory.</p><h3>Breitbart is aggressively mislabeling DIRECT inventory, pooling money and data into Dark Pool Sales Houses</h3><p>During this research, which lasted from early June 2020 through July 15th, 2020, the Breitbart.com/ads.txt file has not seen any changes (or at least the tool being used to track these changes hasn’t flagged any).</p><p>During this period, there were a total of 1,112 total ads.txt records, with 233 of the account-bidding IDs used in online advertising bidding having the “DIRECT” label and 879 having the “RESELLER” label.</p><p>In reality, it seems that <strong><em>only 16 of these DIRECT labeled account-bidding IDs are actually directly controlled by Breitbart</em></strong>, and without being cloned on additional websites.</p><p>Another way to put this, <strong>1,096 of the total 1,112 account-bidding IDs</strong> in the Breitbart ads.txt record are either Resellers and/or part of a cross-domain Dark Pool Sales House.</p><p>Breitbart currently mislabels their inventory via the DIRECT/RESELLER account-bidding IDs— and the site claims 20.95% of the advertising inventory as DIRECT, when in reality only 1.44% of their inventory is DIRECT, with the difference being due to Dark Pool Sales Houses sharing DIRECT labels across businesses and collaborating on the money and user data that flows through those shared account-bidding ID.</p><p>Another way to put this: 98.56% of the inventory on Breitbart.com is either RESELLER inventory or Dark Pool Sales House cross-business inventory mislabeled as DIRECT.</p><h3>Advertisers can’t tell if they’ve been funding Breitbart</h3><p>If you are NOT trying to show ads on Breitbart, or share your money and data with Breitbart’s Dark Pool Sales House partners, this means that there are literally hundreds, if not thousands, of websites that you would need to block.</p><p>In order to not block the websites completely (but still avoid Breitbart’s Dark Pool Sales Houses), you’ll need to block specific account-bidding IDs found in the ads.txt from those websites. This can get complicated…</p><p><strong>We’ll be sharing a list of many of those sites below, and details on how to find them…but this process isn’t cut and dry and many advertising networks need to provide better tools for blocking specific account-bidding IDs across any website where they could show up in an auction, essentially ~dynamic ads.txt account-bidding ID block lists.</strong></p><p>If you ARE trying to show ads on Breitbart (ugh), that’s a bold choice, but if you really must do this, there are only a few account-bidding IDs that are truly DIRECT, and you should direct your spend towards these exchanges &amp; account-bidding IDs on Breitbart.com:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*itTC4D8AdJEGLGm7VVroJw.jpeg" /><figcaption>From Breitbart.com/ads.txt</figcaption></figure><blockquote>addroplet.com, 142, DIRECT</blockquote><blockquote>appnexus.com, 11273, DIRECT, f5ab79cb980f11d1</blockquote><blockquote>appnexus.com, 3936, DIRECT, f5ab79cb980f11d1</blockquote><blockquote>brightcom.com, 24535, DIRECT</blockquote><blockquote>facebook.com, 1243316582352556, DIRECT</blockquote><blockquote>facebook.com, 605216793016672, DIRECT</blockquote><blockquote>google.com, pub-9229289037503472, DIRECT, f08c47fec0942fa0</blockquote><blockquote>google.com, pub-9375184944747545, DIRECT, f08c47fec0942fa0</blockquote><blockquote>infolinks.com, 3192217, DIRECT</blockquote><blockquote>nextmillennium.io, 6010, DIRECT, 65bd090fa4a1e3d6</blockquote><blockquote>pixfuture.com, 3128, DIRECT</blockquote><blockquote>revcontent.com, 80018, DIRECT</blockquote><blockquote>revcontent.com, 80163, DIRECT</blockquote><blockquote>synacor.com, 82063, DIRECT, e108f11b2cdf7d5b</blockquote><blockquote>zedo.com, 1792, DIRECT</blockquote><blockquote>zedo.com, 1793, DIRECT</blockquote><p>To make matters worse, out of that list above, if you are still purposefully trying to buy ads on Breitbart.com, several of the exchanges listed above have significantly more records in the Breitbart ads.txt file and some of those records are part of Dark Pool Sales Houses — so you will need to be able to exclude certain account-bidding IDs for each advertising exchange, or otherwise you will need to exclude the full exchange to ensure you don’t end up bidding on some account-bidding ID for a specific exchange that is actually part of a RESELLER or Dark Pool Sales House inventory.</p><p>So the “final list” of ~safe account-bidding IDs that have no records on that exchange in the Breitbart ads.txt record that are being mislabeled in any way, becomes:</p><blockquote>addroplet.com, 142, DIRECT</blockquote><blockquote>facebook.com, 1243316582352556, DIRECT</blockquote><blockquote>facebook.com, 605216793016672, DIRECT</blockquote><blockquote>nextmillennium.io, 6010, DIRECT, 65bd090fa4a1e3d6</blockquote><blockquote>zedo.com, 1792, DIRECT</blockquote><blockquote>zedo.com, 1793, DIRECT</blockquote><p>If I were helping someone purposefully target ads on Breitbart.com (ugh), there are only 6 account-bidding IDs that can currently be safely used on an exchange bidding platform without ID-level filters.</p><p>Some notes on the<strong> </strong>6 “clean” DIRECT records (from 4 unique companies) out of 1,112 total records on Breitbart.com/ads.txt:</p><ul><li>It’s odd that Facebook is on this list, but Facebook also has slightly more strict policies with shared accounts so it’s possible that Facebook is a known “Not Safe for Sales Houses Advertising Network.”</li><li>Addroplet.com is seemingly new/small, so their policies are unclear.</li><li>NextMillennium.io seems to pride itself on quality and is possibly a “Not Safe for Sales Houses Advertising Network” although they do seem to have some mislabeled DIRECT inventory with records like “nextmillennium.io, 5534, DIRECT, 65bd090fa4a1e3d6” showing up on Adweek.com, Gothamist.com, TalkingPointsmemo.com, and several other sites.</li><li>Zedo.com is also possibly a “Not Safe for Sales Houses Advertising Network” — one of their investors is the legendary<a href="https://en.wikipedia.org/wiki/Esther_Dyson"> Esther Dyson</a>, who guides countless ethical technology and society projects.</li></ul><p>For almost all advertising exchange companies listed here, it’s unclear what their Dark Pool Sales House policies are. Some probably explicitly ban shared Dark Pool Sales House accounts and ads.txt labels, whereas others may just not address it or police it when the DIRECT labels are misused.</p><p>If any organizations listed in this research wants to provide clarity or additional details on their policies, I’ll update sections as-needed with the new information and note the changes.</p><p>As it stands today, if you really wanted to buy ads on Breitbart, as much as I can discourage that practice, there are only 4 companies that I would argue can guarantee some semblance of “quality” due to Breitbart’s mislabeling of DIRECT inventory across countless advertising exchanges.</p><h3>Breitbart doesn’t label their Dark Pool Sales Houses — but RT.com does!</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Vliqc0OGuI9EEXVLyhdt5g.jpeg" /><figcaption>Lots of records, lots of stories @ <a href="https://www.rt.com/ads.txt">https://www.rt.com/ads.txt</a></figcaption></figure><p>The ads.txt standards don’t specifically require labeling Sales Houses because the “RESELLER” label was supposed to take care of that. But due to organizations like Breitbart and RT.com mislabeling their RESELLER inventory as DIRECT, and other operational imperatives with resellers, some organizations have taken to organizing their ads.txt file in Sales Houses with literal company names/URls who are providing the service.</p><p>RT.com does this Sales House labeling with a coded “#Block 1, 2, 3” reference. There are 13 Blocks/Sales Houses in the<a href="https://www.rt.com/ads.txt"> RT ads.txt file</a>, along with several other sections that appear to be unique Sales Houses.</p><p>Some publishers do similar labeling to RT.com, and some even indicate the date that a Sales House was added to the files. There are no official standards for this Sales House labeling and most ads.txt parsers would likely ignore these details or be unable to ingest and parse the notes at-scale.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YKR3u9V-U_BrVgizJoe9Yw.jpeg" /><figcaption>Fox9.com/ads.txt has some interesting Sales House names</figcaption></figure><h3>Many companies are helping publishers create Dark Pool Sales Houses…or looking the other way as it happens…</h3><p>As mentioned throughout this research, huge portions of the Breitbart DIRECT inventory is currently mislabeled, and is actually RESELLER inventory and/or part of a Dark Pool Sales House. Here’s how this works…</p><p>Here’s what a account-bidding ID from Breitbart.com/ads.txt looks like:</p><blockquote><strong>33across.com, 0010b00001shFQpAAM, DIRECT, bbea06d9c4d2853c</strong></blockquote><p>This record can be broken down into parts:</p><blockquote>Advertising Exchange: <strong>33across.com</strong></blockquote><blockquote>AccountID: <strong>0010b00001shFQpAAM</strong></blockquote><blockquote>Type of Inventory Being Alleged: <strong>DIRECT</strong></blockquote><blockquote>Anti-Fraud Tag: <strong>bbea06d9c4d2853c</strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AdCtvjjypIVKTs2HbZ3YqQ.jpeg" /><figcaption>At least some Saambaa clients mislabel DIRECT inventory, including Breitbart.com <a href="https://go.saambaa.com/for-publishers/">https://go.saambaa.com/for-publishers/</a></figcaption></figure><p>This <strong>exact account-bidding ID, including with the “DIRECT” label</strong>, can be found on at least <strong><em>34 websites</em></strong>. This is mislabeled advertising inventory.</p><p>And thanks to some sites organizing and labeling their Sales Houses, including Fox9.com, we can assume this account-bidding IDs is part of a Dark Pool Sales House facilitated by a company called Saambaa.</p><p>Saambaa CEO Matt Voigt reached out after publication to provide additional context to their work, stating:</p><blockquote>“Specifically on the ads.txt Direct issue: We run an events discovery module which renders itself on the page as a Javescript widget. The module is our own content and own experience, which is why we represent the inventory as our own inventory. We do not manage inventory for publishers, or run any kind of publisher ad network.</blockquote><blockquote>We don’t work with Breitbart, but since any publisher can copy and paste our ads.txt up, we don’t control what they have up. It’s an unfortunate reality of ads.txt that there is no control of it for us. The Breitbart ad inventory is managed by a third party ad management company. We work with them on other sites and they have grouped our ads.txt as part of their larger assembly of ad buyers on Breitbart.” — Matt Voigt, Saambaa.com CEO</blockquote><p>For it’s credit, <a href="https://go.saambaa.com/">Saamba</a> is pretty clear about what services they provide, and many organizations who currently facilitate Dark Pool Sales Houses through mislabeled DIRECT inventory, use language about publisher networks and predictable revenue that are valid products. Most of these publisher services should be fine when the shared inventory is labeled as RESELLER inventory, but it becomes a problem when this shared inventory is being labeled as DIRECT.</p><p>At least some of the Saambaa clients mislabel DIRECT inventory, including Breitbart.com<a href="https://go.saambaa.com/for-publishers/"> https://go.saambaa.com/for-publishers/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Um3m5McakCtAIYcoubTtfA.jpeg" /><figcaption>The Progressive “feel good” clickbait shop at Upworthy.com have been collaborating with Breitbart through their Dark Pool Sales House shared bidding IDs — money and data pooling between the organizations.</figcaption></figure><p>By reviewing just a few of the 34 websites that share this DIRECT label, including Upworthy.com, it’s relatively easy to confirm that this shared DIRECT ID is definitely part of the Saambaa Sales House…</p><p>The screen shot to the left is from Upworthy’s ads.txt file — they are among the websites who indicate the Sales House names, as well as the date they were added into their ads.txt records. So we can assume that this Saambaa inventory that overlaps with Breitbart was added on “10/28&quot;, which is likely October 28th, 2019 based on the known timing of deployments as ads.txt has ramped up.</p><h3>Here are 34 websites that share DIRECT ID’s with Breitbart</h3><p><em>There are many other websites, but these are just a few!</em></p><p>Through the Saambaa Dark Pool Sales House, there are 34 websites (including Breitbart)) sharing just one of the Breitbart account-bidding IDs in their ads.txt records. Many of these websites will come as a surprise. But other domains are clearly part of an overlapping network with Breitbart — certain websites show up in Sales House after Sales House with Breitbart, and a pattern emerges that seems to show certain conservative, Republican, and far-right websites are coordinating to deceive advertising buyers by mislabeling their RESELLER inventory as DIRECT.</p><p>Here are the 34 sites sharing one mislabeled DIRECT account-bidding IDs:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Tor38ZaKfa-ExLKsswgnA.jpeg" /><figcaption>Vimeo is collaborating with Breitbart.com, DrudgeReport.com and other members of this Dark Pool Sales House who are mislabeling this RESELLER inventory as DIRECT @ <a href="https://vimeo.com/ads.txt">https://vimeo.com/ads.txt</a></figcaption></figure><p>Breitbart.com</p><p>DrudgeReport.com</p><p>fox9.com</p><p>ReviewJournal.com</p><p>DallasNews.com</p><p>Upworthy.com</p><p>UtahRealEstate.com</p><p>Vimeo.com</p><p>CavsNation.com</p><p>Stocktwits.com</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*azVr2AAgw8uqjmcC8AW5Mw.jpeg" /><figcaption>Major League Baseball? More Like Major League Dark Pool Sales House @ <a href="https://www.mlb.com/ads.txt">https://www.mlb.com/ads.txt</a></figcaption></figure><p>BusinessPress.Vegas</p><p>ClutchPoints.com</p><p>adn.com</p><p>Weatherbug.com</p><p>LeetTemplates.com</p><p>WowHead.com</p><p>InForum.com</p><p>Champion.gg</p><p>ChicagoReader.com</p><p>YachtWorld.com</p><p>HighSchoolLot.com</p><p>mlb.com</p><p>mother.ly</p><p>Allmusic.com</p><p>Good.is</p><p>Hotdeals.com</p><p>RefDesk.com</p><p>Rap-Up.com</p><p>PlymouthHerald.co.uk</p><p>BirminghamMail.co.uk</p><p>Liverpoolecho.co.uk</p><p>Mirror.co.uk</p><p>Examiner.co.uk</p><p>Publir.com (via the Publisher @ <a href="https://a.publir.com/ads-txt/9/ads.txt">https://a.publir.com/ads-txt/9/ads.txt</a>)</p><p><strong>Publir.com is a company associated with a few of the Breitbart Dark Pool Sales Houses so it deserves a brief mention.</strong> The company describes itself as a business who helps “optimize revenues in the programmatic advertising market” — they provide several other services as well.</p><p>Some of the Publir publishers host their ads.txt file on the Publir domain, and some are part of Dark Pool Sales Houses with Breitbart. In this instance, many of the publications involved in the Sales House are listed on the Publir website, and they include The Federalist, PJ Media, Real Clear Politics and many other names familiar to U.S. politics readers — but you’ll also find other odd and low quality websites mixed into many of these Dark Pool Sales Houses, a tactic which raises questions about potentially unique ad fraud schemes.</p><p>Here’s one interesting and massive ads.txt record from Publir to explore (this record and additional ads.txt files will potentially be covered in future research) @ <a href="https://a.publir.com/ads-txt/9/ads.txt">https://a.publir.com/ads-txt/9/ads.txt</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ClAleD5NWviEZgmdF1u1VQ.jpeg" /><figcaption>Many of the Publir publishers host their ads.txt file on the Publir domain, and are part of Dark Pool Sales Houses with Breitbart; One Record @ <a href="https://a.publir.com/ads-txt/9/ads.txt">https://a.publir.com/ads-txt/9/ads.txt</a></figcaption></figure><h3>Dark Pool Sales Houses need more scrutiny from the advertising industry and regulatory oversight</h3><p>Many of the Breitbart.com Dark Pool Sales Houses are labeled on ads.txt files hosted on other websites — so the names of the companies can be found, parsed and flagged. Some of these organizations are violating ethical standards, industry protocols endorsed by the IAB and Google, and likely legal frameworks in numerous jurisdictions, particularly due to their work as data amalgamators without registering as data brokers.</p><p><strong>Mislabeled advertising inventory is very similar to mislabeled securities.</strong> If a buyer doesn’t know what they are buying, that’s a buyer risk. But if the buyer is being told that an asset is of an asset category of greater value than the asset they are actually acquiring, there is an important line being crossed when the seller is purposefully mislabeling that asset, and coordinating with multiple other businesses to mislabel the assets.</p><p>Some of the Breitbart.com ads.txt records mislabeling their DIRECT inventory are found on just a few websites, and aren’t clearly labeled. Some of the mislabeled DIRECT inventory is found on hundreds of websites, with wildly different quality standards and audience profiles across the sites. And some of the mislabeled DIRECT inventory raises concerns about potential ad fraud or inauthentic coordinated activity — or at least significant questions about revenue sharing agreements and cross-publisher agreements that should be disclosed through proper RESELLER inventory labels.</p><p>The ads.txt record “<strong>appnexus.com, 1899, DIRECT, f5ab79cb980f11d1</strong>” for instance, is found on Breitbart.com, but also found on 3 other domains including:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KvYXLWh5AauBqM6F25urMw.jpeg" /><figcaption>This is a fun website to try and figure out the ownership @ <a href="https://misspennystocks.com/about/">https://misspennystocks.com/about/</a></figcaption></figure><ul><li>Breitbart.com</li><li>MissPennyStocks.com</li><li>Tribunnews.com (Indonesian publisher, Subsidiary of<a href="https://www.kgmedia.id/"> https://www.kgmedia.id/</a>)</li></ul><p>The ads.txt record “<strong>avantisvideo.com, 7729, DIRECT</strong>” is another found on Breitbart.com, but also found on several other domains, creating a Dark Pool Sales House of mislabeled DIRECT inventory.</p><p>The websites associated with this account-bidding ID are:</p><ul><li>Breitbart.com</li><li>Takimag.com (Screen shot of columnists below)</li><li>RicksFreeAutoRepairAdvice.com (This site has a huge amount of ads, with lots of high paying video ads and multimedia ads . They also host their ads.txt with the Conservative Publisher Dark Pool Sales House at Publir.com</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FQl6FixtspxPemSJhwWz1Q.jpeg" /><figcaption>Takimag.com, who mislabels RESELLER inventory via a shared account-bidding IDs with Breitbart and “RicksFreeAutoRepairAdvice.com” has Republican columnists including Ann Coulter and Pat Buchanan</figcaption></figure><p>As mentioned throughout this research, there are a huge number of account-bidding IDs being mislabeled as DIRECT on Breitbart.com, and there could be dozens of interesting Dark Pool Sales Houses that are created, or potential low quality spam sites within the networks. We strongly encourage anyone involved in online buying or looking to navigate the complex world of online advertising Dark Pool Sales Houses, to start searching the Breitbart account-bidding IDs labeled as DIRECT, literally searching the full records in quotes on Google, and you’ll find most or all of the websites and direct links to those ads.txt records to confirm.</p><h3>Conclusion and Next Steps</h3><p>Ads.txt is only three years old, but instead of reducing ad fraud, it has created whole new vectors of mislabeled inventory being abused by Dark Pool Sales Houses and facilitated, or ignored, by advertising companies across the world.</p><p>When ads.txt was first debated, the IAB and other advertising organizations knew that a Sales House would exist — which is why the entire concept of a RESELLER label was created. There is nothing wrong with a properly labeled RESELLER account-bidding ID for a Sales House within an ads.txt record — entire sectors of the advertising industry are structured to provide this type of audience amalgamation service.</p><p>But an important line is crossed when account-bidding IDs are labeled as DIRECT and cloned across multiple websites owned by different companies. This does not align with the intent of this label as defined by the original standards, nor does it align with how the label is used by legitimate publishers. This mislabeling is by many standards a form of advertising fraud — and it’s a breach of trust for buyers who are trying to navigate the new and complex ads.txt advertising standard.</p><p>In the near future, advertising exchanges should strongly clarify their policies around Dark Pool Sales Houses — that is — mislabeled DIRECT inventory across multiple websites owned by different companies. And hopefully, buyers who participated and won auctions from these IDs, should get refunds on the lower quality inventory they potentially received.</p><p>There are still many important discussions to have around Dark Pool Sales Houses, and potentially new legislation that should be crafted to articulate rules and regulations for these markets. There is likely some additional research that will be released on this topic, and hopefully some updates from advertising exchanges who clarify their policies.</p><h3>Online Advertising Brand Problems Require Businesses to “Check Your Ads” — and Why Co-Founders Nandini Jammi and Claire Atkin Understand Brand Risk in the Online Ads Space in the Era of Fake News, Disinformation, and Hate Speech</h3><p>This research is being published in collaboration with Nandini Jammi and Claire Atkin, co-founders of Check My Ads (<a href="https://www.checkmyads.org/">https://www.checkmyads.org/</a>) and publishers of <a href="https://branded.substack.com/">the must-subscribe newsletter BRANDED</a>.</p><p>Their work is a one stop shop for all breaking news in brand safety, online advertising changes and how to effectively navigate the complicated world of online ads and fake news. Their new company Check My Ads gets an extremely strong (unpaid) endorsement from me — and anyone buying online ads in any significant amount/frequency needs to strongly consider reaching out to them to understand their process and products.</p><p>Nandini Jammi is also the<a href="https://www.nytimes.com/2018/07/20/business/media/sleeping-giants-breitbart-twitter.html"> co-founder of Sleeping Giants</a>, which until this past month I had no clue was a title that was in any form of dispute, particularly some horrific dispute from the white man who co-founded the organization with her. If you haven’t read her important story, please take the time to read “<a href="https://medium.com/@nandoodles/im-leaving-sleeping-giants-but-not-because-i-want-to-d9c4f488642">I’m leaving Sleeping Giants, but not because I want to</a>”</p><p>This open letter from Nandini was published while we were working on this ads.txt research, and it was an unfortunate reminder that activists can’t ignore problematic peers, and we must ask the tough questions when we see women and POC being sidelined or not properly given the exposure for their work. Personally, I know I can always do better here, and I think every white man should be regularly assessing their behavior and conduct to elevate any peers that aren’t getting their proper credit.</p><p>Briefly about Sleeping Giants and the relationship to this research… from my perspective, I had read the 2018 NYTimes article and had engaged with the group back in 2016, and I always thought there were 2 co-founders — which there obviously are.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bh4pDoETTHf6is7vBAHRaA.jpeg" /></figure><p>But back in 2016, I was skeptical of the Sleeping Giants strategy (I was wrong), but also<a href="https://twitter.com/thezedwards/status/812684039014453248"> encouraged them via Twitter</a> to parse the advertising network ID available in the source code of websites to find more of the networks associated with Breitbart. SG never reached out to me and I dropped the issue, but this concept has been in the back of my head for a few years (and used by me to analyze tons of other networks) — and during that time, ads.txt launched and became the core way for sharing account-bidding IDs across businesses.</p><p>In June of this year, after having connected with Claire earlier in the Spring and seeing their BRANDED newsletter start to tackle important online advertising and brand safety topics (and rack up wins!), I reached out directly to Nandini to talk about this account-bidding ID-mapping concept I first raised in 2016. I had seen some of Nandini and Claire’s recent brand protection wins, and one of their tweets reminded me of the idea again to parse account-bidding IDs to map Breitbart partnerships. We started talking and this research is the result.</p><p>While some of the research here into the specific ads.txt Sales Houses, and some of the technical / market conclusions here are my personal opinion/analysis, this research wouldn’t be possible with the work of Nandini and Claire, and I’m confident that they will both prove anyone wrong who doubts or discounts their power to change how brands and advertisers approach online ads.</p><p>The three of us spoke throughout the research and their insights into buying practices, brand safety strategies, and their understanding of risk mitigation in online buying is at the top of the game, and a real inspiration for data supply chain researchers like myself.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fvPQf-pAzDBHhlYHFXbReA.jpeg" /><figcaption>The author (Zach) with Andrew Breitbart in 2011 at a digital conference.</figcaption></figure><p>Personally, I’ve worked with some of the biggest and best digital ad buyers in the world, and there are less than half a dozen that I would trust with friends/family/client budgets and optimization strategies. Nandini and Claire have my strongest recommendation, and can’t wait to see their careers continue to shape the world and internet we live and conduct business in.</p><p>Nandini and Claire have also published a detailed summary and action guide in their newsletter about this Ads.txt research —<a href="https://branded.substack.com/p/so-thats-how-breitbart-is-still-making"><strong><em>you should click here</em></strong></a><strong><em> and open their work in another tab and read it right now.</em></strong></p><p>Please don’t hesitate to ping any of us on Twitter (@<a href="https://twitter.com/nandoodles">nandoodles</a>, @<a href="https://twitter.com/catthekin">catthekin</a>, @<a href="https://twitter.com/thezedwards">thezedwards</a>) with any questions or concerns.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6e7e3b5c3318" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[July 2020 Compromised PaF Subdomains (mostly via Microsoft Azure)]]></title>
            <link>https://medium.com/@thezedwards/july-2020-compromised-paf-subdomains-mostly-via-microsoft-azure-5834ae22733a?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/5834ae22733a</guid>
            <category><![CDATA[malware]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[subdomain-takeover]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[dns]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Wed, 08 Jul 2020 01:19:08 GMT</pubDate>
            <atom:updated>2020-07-10T23:01:30.335Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_Oy_zgO9fXoa9TqYz20kKA.jpeg" /><figcaption>This is what many of the compromised subdomain homepages look like — “coming soon” type pages in different languages…</figcaption></figure><p>Continued from Twitter… please read this thread before engaging… Crowdsourcing research project ahead!</p><h3></h3><p></p><p>Please be extremely careful with the subdomains listed below — many of these are still compromised. If you find a compromised subdomain, please consider reaching out to anyone at that organization who could take down the subdomain.</p><p>Ping me @thezedwards on Twitter and i’ll help with whatever I can. If you are an organization who doesn’t want to be on this list, ping me and i’ll see what I can do.</p><p>Please consider this a public service — I’m trying to get get these subdomains fixed as soon as possible, provide a resource for security organizations to find and report more of these compromises, and as of publishing time, many of the organizations who have compromised subdomains listed below, haven’t taken action, which puts their users at risk for compromises.</p><p><a href="https://www.simoahava.com/web-development/whats-in-a-cname/">Here’s a solid article on the technical realities that exist when a subdomain is compromised via this type of CNAME mapping…</a></p><p><strong><em>All organizations should make it a top priority to clean up these compromised subdomains!</em></strong></p><h3>Current List of July 2020 Compromised PaF Subdomains</h3><p>360stage.stahls.com</p><p>analytics.glamst.com</p><p>b2btdc.pandora.net</p><p>beta-invited.slh.com</p><p>ccc.blockshipping.io</p><p>champions-d-content.generalmills.com</p><p>chat.celcom.com.my</p><p>cine.naturgy.es</p><p>cloudpilotsg.cloudatlasinc.com</p><p>costwell.chevron.com</p><p>demo.booktrack.com</p><p>drweb.commscope.com</p><p>elevate17.bittitan.com</p><p>eperfectlaunchdev.optum.com</p><p>farmtoschoolmap.georgiaorganics.org</p><p>findyourstyle.fisherpaykel.com</p><p>game.autoshow.ca</p><p>gifts-uat.unrefugees.org.au</p><p>greathallcontacts.flydenver.com</p><p>hippotalk.total.com</p><p>devagileblog.acuitybrandslighting.net</p><p>devoddsapi.wallstreetenglish.com</p><p>dvsm-uat.gsk.com</p><p>es-stgics-avm.jll.com</p><p>returns.americas.pandora.net</p><p><a href="http://www.iknow.dr.cch.com.au/">www.iknow.dr.cch.com.au</a></p><p>m.macaronigrill.com</p><p>map.carlgross.com</p><p>mobile-beacons.clearchannel.co.uk</p><p>mobile.hullcitytigers.com</p><p>mobile.stratasys.com</p><p>old.deleteagency.com</p><p>oneanalytics.capita-one.co.uk</p><p>partners.honeygroup.co.uk</p><p>pay.willassociates.co.uk</p><p>peerwatch.complianceweek.com</p><p>portfolio.theglobalfund.org</p><p>ppkpi.cbre.com</p><p>prdmarep.udtrucks.com</p><p>prod.vallen.ca</p><p>production.go-dove.com</p><p>pwcs-grants.pwcs.edu</p><p>recommendations.govx.com</p><p>secure.openenergymarket.com</p><p>sfgateway-prod-east-api.carmax.com</p><p>sportsfirstaid.redcross.org.uk</p><p>sso-api-poc.mybswhealth.com</p><p>stage.cleanwithkeystone.com</p><p>staging.auth.idahopower.com</p><p>stagingcms.johnsoncontrols.com</p><p>storetool.albertsons.com</p><p>storetool2.albertsons.com</p><p>ticari.mercedes-benz.com.tr</p><p>voar-tracking.volvo.com</p><p>wiki.gibson.com</p><p>wine.mydexrewards.com</p><p>wisent.mitt.ru</p><p>wisent.mosbuild.com</p><p><a href="http://www.app.ahvoila.com/">www.app.ahvoila.com</a></p><p><a href="http://www.sensformer.cloud.siemens.com/">www.sensformer.cloud.siemens.com</a></p><p>xlcatlin.leopard-np.swissre.com</p><p>trace.accenthealth.com</p><p>linode.hki.org</p><p>advanced.core.freeflow.xerox.com</p><p>nucleus.robomateplus.com</p><p>quantumleap.pason.com</p><p>un1cdp01.uno.adt.bms.com</p><p>booking.ramadadowntowndubai.com</p><p>login.ec.co</p><p>b2b.absoluteboardco.com</p><p>pfp-int.az-bots-gre-projets.viseo.com</p><p>dashboard.adsninja.com</p><p>scm.ordermanagement-test.maersk.com</p><p>maps.foundationcenter.org</p><p><a href="http://www.thevillagesatpinevalley.com/">www.thevillagesatpinevalley.com</a></p><p>itpolicies.ycp.edu</p><p><a href="http://www.summary.batransfer.com/">www.summary.batransfer.com</a></p><p>chat.fnv.nl</p><p>dev-cd-infocenter.ryder.com</p><p>myob-multi-dc-sit-singapo-cfs-v1.myob.com</p><p><a href="http://www.loveisajourney.proflowers.com/">www.loveisajourney.proflowers.com</a></p><p><a href="http://www.satisfaction.darty.com/">www.satisfaction.darty.com</a></p><p>internationalservicesstage.rrd.com</p><p>detectionapp.3m.com</p><p>pspapimgmt-test.premera.com</p><p>b2bapi-service-acc.snelstart.nl</p><p>aem.herbalife.com</p><p>v2.basic.net</p><p>usersapipre.vertele.eldiario.es</p><p>onespie.spie.de</p><p>applications.wirralccg.nhs.uk</p><p>beta.pksinvest.com</p><p>site.chopup.me</p><p>tevatogostgrw.tevapharm.com</p><p>thweb-azure.teknikhuset.se</p><p>nexarc1service.kemin.com</p><p>ohmy.disneylatino.com</p><p>sccmclouddp.providence.org</p><p>sitgbapi.globalblue.com</p><p>qcsampler.genpact.com</p><p>geaux.lsufoundation.org</p><p>apps.technologydev.ihs.com</p><p>storelocator.dtc.newbalance.com</p><p>football.swisslife.ch</p><p>inflightentertainment.sas.no</p><p>connected.virginaustralia.com</p><p>inhabit-portal.arkadium.com</p><p>beta.auic.org</p><p>iot-accelerator-dev2.ddm.iot-accelerator.ericsson.net</p><p>api.elfcosmetics.com</p><p>accessderm.aad.org</p><p>cmclouddpsgsin.autodesk.com</p><p>search.us.epg.toshiba.com</p><p>uoncmgtst.newcastle.edu.au</p><p>blog.codercamps.com</p><p>members.i.playboy.com</p><p>zew-api.travelport.com</p><p>aicpasccm.aicpa.org</p><p>smartusw-sts.gep.com</p><p>hatchery.entrepreneurial-spark.com</p><p>bmsazure.elas.uk.com</p><p>referencement.levio.ca</p><p>iq.aecom.com</p><p>a.eage.org</p><p>poc9.icertis.com</p><p>uat.ovhq.msc.com</p><p>tibco-service-dev.usga.org</p><p>icqa.skillsinsight.honeywell.com</p><p>members.ussvi.org</p><p>associate.myfortisonline.com</p><p>acdadmin-tng.aia.org</p><p>demo06.mediusflow.com</p><p>myaccount.scottish-enterprise.com</p><p>nw-b.ecolab.com</p><p>members.dotnetfoundation.org</p><p>automation.pg.com</p><p>mclambda-devtest.cpsextsandbox.mayo.edu</p><p>dev.forsyteit.com</p><p>testazure.drivetime.com</p><p>ve-service.genecards.org</p><p>ahbeardweb.microsoftcrmportals.com</p><p>wordpress-itec.azurewebsites.net</p><p>qa.api.sapaccess.warnerbros.com</p><p>stpaul.partnerinhousing.com</p><p>dashboard.boostup.com</p><p>docs.cms.orckestra.com</p><p>ecmcmg.broadinstitute.org</p><p>cms.facilitiessurvey.com</p><p>dev1.mdlive.com</p><p>aauw-ampostdoc.scholarsapply.org</p><p>tge.tradeglobal.com</p><p>mobile.apply4housing.com</p><p>my.disciplesmade.com</p><p>quote.model.healthmarkets.com</p><p>dev.connectedservices.emerson.com</p><p>connect.atslab.com</p><p>training.trin.net</p><p>stgwww.ispeedyloans.com</p><p>mossupport.mcd.com</p><p>prepd-sitecore.solr.arm.com</p><p>spaspera.cloud.jci.com</p><p>uopxcmg2.phoenix.edu</p><p>staging-consulting-covid19.euromonitor.com</p><p>mail.somersetcm.com</p><p>dev.salesforce.integration.plex.com</p><p>https:full-service-suite.ch</p><p>cart.perseusacademic.com</p><p>testwebservices.hawaiianairlines.com</p><p>timesheets.cfed.org</p><p>library.inthehand.com</p><p><a href="http://www.rmspecialstamps.com/">www.rmspecialstamps.com</a></p><p>sessions.digitalwpc.com</p><p>staging.ecofastensolar.com</p><p>innovapulse.innovasi.com</p><p>uk.ziraat.turkline.com</p><p>rldp.redlobster.com</p><p>test-cbreitp.intrepid.cbre.com</p><p>go.daymarksi.com</p><p>test.lark-it.com</p><p>dynamicsac.perficient.com</p><p>voyager-dev.kindred.com</p><p>acsdonateadmintrain.cancer.org</p><p>prixmnbawards.musicnb.org</p><p>nlgsccmconnect1.nationallife.com</p><p>create.cakesbyron.com</p><p><a href="http://www.mitanorifusa.com/">www.mitanorifusa.com</a></p><p>dev-oms-logistics.pvh.com</p><p>sts.hgem.com</p><p>gettyclouddp1.gettyimages.com</p><p>training.iverson.com.my</p><p>secure.web.powerapps.com</p><p>cb.us.stg.cloud.im</p><p>press.desigual.com</p><p>architectuur.cibg.nl</p><p>myusa.veinteractive.com</p><p>qa.boh.com</p><p>xlcatlin.leopard-np.swissre.com</p><p>blog.washingtonstem.org</p><p>apps.invictusgames2017.com</p><p>test.scandichotels.de</p><p>sccm-dp.acuitybrands.com</p><p>fnmaxcmgdp.fanniemae.com</p><p>survive.infocomm.org</p><p>op.elfcosmetics.com</p><p>leprdsccmdistpteuwest.lincolnelectric.com</p><p>cms.answersmediainc.com</p><p>cloudsolvportal.synnex.com</p><p>uatstandby-www.cushmanwakefield.com</p><p>emergencyresponse.bristowgroup.com</p><p>thor.mdlive.com</p><p>clouddp01.lamresearch.com</p><p>autoattendantservicesqa.incontact.com</p><p>microsoft.icertis.com</p><p>devpmforecaster.cbre.com</p><p>tastings.neudesic.com</p><p>b2bws.julian-fashion.com</p><p>apimcustomapi-dev.azure.chevron.com</p><p>analytics.donorperfect.net</p><p>ecom-qa-nl.bambonature.com</p><p>s-sccmdp-cloud01.loandepot.com</p><p>tra.g4s.com</p><p>remote.packtech.dk</p><p>qaappcenterng.deloitteresources.com</p><p>apps.fullertonhealth.com</p><p>smoke1.remix3d.com</p><p>onesiteportal-stage.rrd.com</p><p>cdn02.empiretoday.com</p><p>easishare.bruker.com</p><p>football.swisslife.ch</p><p>seminario.iipe.unesco.org</p><p>cityofcalgarycmg.calgary.ca</p><p>dailysales.brownjordan.com</p><p>staging.capturetech.com</p><p>media.antenna.gr</p><p>doc.bootes.co</p><p>am.us.rothschildandco.com</p><p>candidate.responsivehr.com</p><p>lti.intelequia.com</p><p>api.longbow.bonusxp.com</p><p>tuap.teamusa.org</p><p>rss-prototype.bd.com</p><p>— -</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FxTiTnhXjLyayuvu3hm%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fvulture-xTiTnhXjLyayuvu3hm&amp;image=https%3A%2F%2Fmedia2.giphy.com%2Fmedia%2FxTiTnhXjLyayuvu3hm%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="245" frameborder="0" scrolling="no"><a href="https://medium.com/media/848514d1904c7a8a00d06b3b4d3f3f91/href">https://medium.com/media/848514d1904c7a8a00d06b3b4d3f3f91/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5834ae22733a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Final Statement of Reasons from The California Attorney General for CCPA Raises Important Questions]]></title>
            <link>https://medium.com/golden-data/final-statement-of-reasons-from-the-california-attorney-general-for-ccpa-raises-important-questions-ce63a90d97cc?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/ce63a90d97cc</guid>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[ccpa]]></category>
            <category><![CDATA[california]]></category>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[law]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Wed, 10 Jun 2020 04:03:27 GMT</pubDate>
            <atom:updated>2020-06-23T18:52:53.321Z</atom:updated>
            <content:encoded><![CDATA[<h4>Big Data Organizations &amp; Service Providers have weeks to get ready</h4><p>The California Consumer Privacy Act (CCPA)is going to be enforced starting on July 1, 2020 having gone into effect at the start of 2020 — and new guidance from the California Attorney General should quickly become the focus of any digital organizations with significant amounts of user data.</p><p>This blog post is not meant to be an all-encompassing summary of how to get ready for CCPA or the frameworks for sharing and selling user data — there are far too many complicated aspects, largely due to the fact that most organizations who are large enough to need to comply with CCPA, would also have European users and need to comply with GDPR, the European data privacy law.</p><p>There are several comments in the California Attorney General’s “Final Statement of Reasons” for CCPA that <strong>clarifies important rights and responsibilities under CCPA.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YyzlIte2Ho40rt4THUlKRg.jpeg" /><figcaption>NAI testimony before the U.S. Senate <a href="https://www.commerce.senate.gov/services/files/3AE22675-7424-4CCD-859A-DDBBD3CAA082">can be viewed here</a>.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*h9ASBsw-Etrid8lYzR74NQ.jpeg" /><figcaption>IAB testimony before the U.S. Senate <a href="https://www.commerce.senate.gov/services/files/97962BF5-D0A7-41F4-937D-72D8CD06623E">can be viewed here</a>.</figcaption></figure><p>These details are being released at a time when COVID mobile tracking data has become the newest privacy outrage for users — and several aspects of the guidance reads as a direct rejection of the guidance issued by the online advertising and analytics industry groups <a href="https://www.commerce.senate.gov/services/files/3AE22675-7424-4CCD-859A-DDBBD3CAA082">NAI</a> and <a href="https://www.commerce.senate.gov/services/files/97962BF5-D0A7-41F4-937D-72D8CD06623E">IAB</a>, who previously gave their members a blessing to share/sell COVID mobile tracking data to other businesses, researchers and the government to support the pandemic tracking efforts.</p><p>Both IAB and NAI encouraged members to share any data valuable against fighting COVID in the Senate hearing that was not on video, via their written statements for the hearing “<a href="https://www.commerce.senate.gov/2020/4/enlisting-big-data-in-the-fight-against-coronavirus">Enlisting Big Data in the Fight Against Coronavirus</a>.”</p><p>It’s clear that organizations who buy/sell/share user data, need to get much more serious about user consent, the categories of collection they undertake, and their potential legal exposure from not requesting user consent for a material change in collection purpose — and the CCPA guidance makes it clear that “<strong>simply putting up a new notice on a website</strong> after a consumer has already provided personal information, when that consumer may be unlikely to revisit the website (and even more unlikely to revisit the notice), <strong>is not meaningful consumer notice.”</strong></p><p>If you are a business with significant user data (10+ million consumers in a calendar year), you don’t get to start every month coming up with new monetization strategies for your existing user data without getting permission from users to use their existing data for materially different efforts — and with the new categories of sources being clarified by the CA AG to now include: “Advertising Networks, Internet Service Providers, Data Analytics Providers, Operating Systems and Platforms, Social Networks, and Data Brokers” — things are about to get much more serious for organizations who have treated user consent like a blank check for future user data monetization efforts.</p><p>Brief disclaimer: I’m not a lawyer — i’m a longtime digital strategist who has a significant interest and experience with user data privacy frameworks (i’ve also got my CIPP/US privacy certification from the <a href="https://iapp.org/">IAPP</a>). I’ve been building and optimizing marketing and analytics stacks for 13+ years for politicians, businesses, my own startups, and client projects — with the last ~8 at my firm <a href="https://victorymedium.com/">Victory Medium</a>. I’m on Twitter @ <a href="https://twitter.com/thezedwards">thezedwards</a> for any questions or feedback.</p><h3>Excerpts from the Final Statement of Reasons from the California Attorney General’s Office, with Brief Comments on the Potential Scope / Impacts</h3><p>The PDF for the <a href="https://oag.ca.gov/sites/all/files/agweb/pdfs/privacy/ccpa-fsor.pdf">Final Statement of Reasons can be viewed here</a>. <a href="https://oag.ca.gov/privacy/ccpa">Additional links and CCPA resources</a> can be found at the CA AG’s website.</p><p>Categories of data sources and the types of entities that collect data have been expanded to include and require more specificity, with several new entity definitions including, “ <strong><em>advertising networks, internet service providers, data analytics providers, operating systems and platforms, social networks, and data brokers</em></strong>.”</p><p>From page 1 of The Reasons:</p><blockquote><strong>Subsection (d) </strong>has been modified to provide further guidance and clarification for the definition of “categories of sources,” which is used throughout these regulations. (See Sections 999.301, subd. (q)(3), 999.308, subd. ©(1)(e), 999.313, subd. ©(10)(b).) “Categories of sources” has been clarified to mean “types or groupings of persons or entities” from which a business collects consumers’ personal information, not just “types of entities.” The definition has also been modified to require a business to describe its categories of sources “with enough particularity to provide consumers with a meaningful understanding of the type of person or entity.” The following examples have also been added to the definition: <strong><em>advertising networks, internet service providers, data analytics providers, operating systems and platforms, social networks, and data brokers</em></strong>.</blockquote><h4>A Business Can be Classified as Both a 1st and 3rd Party, Depending on the Context of the Collection</h4><p>Much like under GDPR where an organization can act as both a Data Controller and Data Processor, CCPA now allows an organization to be categorized as both a 1st party and a 3rd party entity whom businesses share personal information, depending on the context of that collection:</p><blockquote><strong>Subsection (e) </strong>has been modified to provide further guidance and clarification for the definition of “categories of third parties,” which is used throughout these regulations. (See Sections 999.301, subd. (q)(5), 999.308, subd. ©(1)(g)(2), 999.313, subd. ©(10)(d).) “Categories of third parties” has been clarified to mean types “or groupings of third parties with whom the business shares” personal information, rather than “types of entities that do not collect personal information directly from consumers.” The definition has also been modified to require a business to describe its categories of third parties “with enough particularity to provide consumers with a meaningful understanding of the type of third party.”</blockquote><blockquote><strong>These modifications are necessary because entities with whom businesses share personal information may also collect personal information directly from consumers in other contexts.</strong> The CCPA’s definition of “third party” excludes <em>the </em>business that collects personal information from consumers, meaning the business that collects a consumer’s personal information in a particular context; it does not exclude <em>all </em>businesses that collect personal information directly from consumers in any context. (Civ. Code, § 1798.140, subd. (w).) <strong>An entity may in some instances be the business that collects personal information from consumers and in other instances a third party that receives personal information collected by another business. </strong>Accordingly, the definition of “categories of third parties” has been modified to clarify this point. These modifications also provide more guidance to businesses concerning the information they are required to provide to consumers, especially when responding to a request to know. This additional guidance benefits consumers by requiring that businesses provide enough information for consumers to understand their data practices. By requiring businesses to describe categories of third parties in a manner that is easily understood by consumers, these modifications implement a performance-based approach. (See Schaub, Center for Plain Language.) It also benefits businesses, particularly smaller businesses that lack privacy resources, by clarifying the information they must provide to consumers.</blockquote><h4>Households get clarifications</h4><p>Throughout CCPA and the guidance from the California Attorney General’s office, there are mentions of “households” — these are groupings of individuals, sometimes related to each other and other times just living together, who may have overlapping data or an interest in restricting access to their data from other members of the household.</p><p>Technically, a household oftentimes shares an IP address range between the members of the household, which can be used as a persistent identifier by advertising and analytics companies.</p><p>Household Data Access &amp; Deletion requests are going to be challenging for some organizations, and several aspects of the CCPA guidance seems to be aimed at discouraging “Verification by IP Address” for Access/Deletion requests — and basically organizations need to not only account for household requests (group requests), but also come up with their own internal Trust &amp; Safety solutions to reduce the likelihood of a household guest, Airbnb renter or some other temporary occupant taking advantage of an unsafe access/deletion process.</p><p>The CCPA guidance on pages 3–4 includes:</p><blockquote><strong>Subsection (k) </strong>was formerly subsection (h) and has been renumbered. The initial proposed definition of “household” has been modified to “a person or group of people who: (1) reside at the same address, (2) share a common device or the same service provided by a business, and (3) are identified by the business as sharing the same group account or unique identifier.” This change was made in response to comments that the initial proposed definition of “a person or group of people occupying a single dwelling” was overly broad. The change is necessary to ensure that the term does not encompass persons with only a transitory relationship to a dwelling or a tenuous connection to another resident. This change will benefit businesses by providing more guidance about which groups of persons to treat as a household and will benefit consumers by ensuring that those who only temporarily occupy a dwelling are not able to access or delete a consumer’s household information.</blockquote><h4>Offline data collection still requires notification before the “point at which” the business collects personal information</h4><p>There are numerous sections of the CCPA guidance that attempt to provide guidance about when a consumer must be notified about the collection of personal information — and one important part of these regulations could basically implode the entire outdoor kiosk/POS mobileID tracking schemes here in California. Many of these outdoor scanners are basically constantly hovering up consumer data, and reselling it for everything from COVID tracking to Online-offline marketing attribution.</p><p>The CCPA regulations now require consumer notification at or before the “point at which” a business collects personal information from a consumer.</p><p>From page 4:</p><blockquote><strong>Subsection (l) </strong>was formerly subsection (i) and has been renumbered. The initial proposed definition of “notice at collection” required notice to be provided to consumers at or before the “time” of collection of personal information. The definition has been modified to state that <strong>the notice must be provided at or before the “point at which” a business collects personal information from a consumer. </strong>This change is necessary to make the definition consistent with the language used in the CCPA. (See Civ. Code, § 1798.100, subd. (b).) This change is also necessary to encompass both temporal proximity, such as in online data captures, <strong>and physical proximity, such as near a cash register at an in-store location where collection is taking place. </strong>This change will benefit businesses by providing further guidance on how to provide notice to consumers and will benefit consumers by making the notice more apparent when personal information is collected.</blockquote><h3>Guidance on Mobile Device Collections Raise Significant Questions for COVID Location Data Sales</h3><p>The timing of when this CCPA guidance was written is important — these opinions were being written while massive amounts of mobile location data from the public was being bought, sold, and shared under the guise of consumer protection, and with the NAI and IAB advertising industry groups both blessing the practice of selling existing user mobile location data to support COVID tracking efforts.</p><p>What the CCPA guidance makes clear, and this should raise red flags for any organizations who took guidance from NAI and IAB on this issue and executed sales of existing user data, is that the CCPA guidance now makes it clear that organizations who provide SDK services to apps, and any app providing data for COVID tracking, need to provide a “‘just-in-time’ notice summarizing those categories of information that a consumer would not reasonably expect to be collected..”</p><blockquote><strong>Subsection (a)(4) </strong>was added to address instances in which a business collects personal information <strong>from a consumer’s mobile device for purposes that the consumer would not reasonably expect. </strong>In those instances, the business must provide a just-in-time notice summarizing those categories of information that a consumer would not reasonably expect to be collected and a link to the full notice at collection. The subsection also includes an example that illustrates this requirement and provides guidance as to what may be considered a purpose that a consumer would not reasonably expect. <strong>This subsection is necessary to provide transparency into business practices that defy consumers’ reasonable expectations, particularly when those uses are not reasonably related to an application’s basic functionality. </strong>The requirement benefits consumers by making notices more conspicuous in instances in which their personal information is being collected for purposes not reasonably expected. Subsection (a)(4) is consistent with the language, intent, and purpose of the CCPA to meaningfully give notice to consumers about what information is collected from and about them and to give them control over how businesses use this information. The CCPA provides the OAG with the authority to adopt regulations as necessary to further the purposes of the CCPA. (Civ. Code, § 1798.185, subd. (b)(2).) Inherent in this authority is the ability to adopt regulations that fill in details not specifically addressed by the CCPA, but fall within the scope of the CCPA. This just-in-time notice allows consumers to make an informed decision about how to interact with the business at or before the point of collection of their information, in furtherance of Civil Code § 1798.100, subdivision (b). The regulation also benefits businesses by providing clear guidance regarding when they must provide a just-in-time notice on a consumer’s mobile device.</blockquote><p>The CCPA guidance goes on to further highlight mobile tracking situations that require unique disclosures, particularly any new use of consumer data that is “materially different than” the original purpose, writing:</p><blockquote>Former subsection (a)(3) has been renumbered and is now <strong>subsection (a)(5)</strong>. <strong>Subsection (a)(5) </strong>concerns restrictions on a business’s use of a consumer’s personal information for purposes other than those disclosed in the notice at collection. These restrictions are necessary because the consumer could have reasonably relied on the notice when interacting with the business and allowing it to collect their personal information. <strong>The subsection has been modified by changing “other than” to “materially different than.” This change was made in response to numerous comments urging that the restrictions be limited to uses that are “materially different”</strong> from those disclosed in the notice and is necessary to make the language of the regulation consistent with privacy best practices. For example, the FTC has long expected that companies should obtain affirmative express consent before using consumer data in a materially different manner than claimed when the data was collected. (See Fed. Trade Com., Protecting Consumer Privacy in an Era of Rapid Change: Recommendations for Businesses and Policymakers, (2012), p. viii, 57–58.) This change benefits businesses because businesses will not be required to inform consumers of immaterial changes. The change also benefits consumers by not overwhelming them with notices for every minor change, which may result in notice fatigue. The subsection also adds the term “previously collected.” <strong>This change is necessary to clarify that the subsection applies when a business seeks to use previously collected personal information for a use that is materially different than what was previously disclosed to the consumer, not for new personal information that it seeks to collect.</strong></blockquote><blockquote>In light of the comments received from the public, the OAG further supplements its statement of reasons in support of subsection (a)(5) as follows. (See ISOR, pp. 8–9.) Subsection (a)(5) is consistent with the language, intent, and purpose of the CCPA to <strong>provide consumers with greater control over their information and meaningful ability to exercise their CCPA rights.</strong></blockquote><p>The CCPA guidance also includes an example scenario where businesses would be required to request consent for a new purpose:</p><blockquote><strong>When businesses change practices midstream, the consumer should have the opportunity to decide whether to agree to the new purpose.</strong> For example, a consumer may be comfortable allowing a business to collect their personal information to use in serving them advertisements for relevant products, but not if the business wants to use the information to conduct psychological experiments. Allowing consumers the opportunity to consent to this further use is consistent with the CCPA’s goal of fairness, choice, and control.</blockquote><p>And it seems that some businesses were advocating to the California Attorney General that they should be able to merely update a notice / TOS / Privacy Policy, and that CCPA doesn’t provide the OAG authority to strengthen requirements and require consumer-consent for a new data collection purpose.</p><p>The AG’s guidance clearly shot down this argument, and the CCPA guidance seems to make it clear that a new purpose (like COVID location data sales using existing mobile data) would not be CCPA compliant and requires a business to request permission to use the existing data for the new purpose:</p><blockquote>Some comments have interpreted Civil Code section 1798.100, subdivision (b), <strong>as only requiring an additional notice and prohibiting a consumer-consent requirement. The OAG disagrees with this interpretation.</strong> As already stated, the CCPA gives the OAG authority to promulgate regulations that further the purposes of the CCPA. Furthermore, requiring explicit consent puts the consumer in the same position they would have been had the material change been disclosed during the consumer’s first engagement with the business. It empowers the consumer to actively choose whether they want to maintain their relationship with the business. <strong>It is necessary to preserve the consumer’s ability to object to the use of their personal information for new purposes, particularly because the business already has their personal information</strong>. Without this requirement, businesses could be incentivized to minimize or delay any updated notice or to otherwise hinder a consumer’s ability to object to the new use. <strong>Thus, comments that propose simply updating an online privacy policy or providing notice without explicit consent for material changes to a business’s use of personal information would not serve the purpose of section 1798.100, subdivision (b).</strong> Such an approach would allow businesses to engage in passive notice updates without allowing consumers any agency to control how their personal information is used. <strong>Furthermore, simply putting up a new notice on a website after a consumer has already provided personal information, when that consumer may be unlikely to revisit the website (and even more unlikely to revisit the notice), is not meaningful consumer notice.</strong></blockquote><p>If you are an organization who has been using existing mobile location data from apps or SDKs and selling/sharing or using that data in any way to support COVID tracking efforts, that seems to be a significant CCPA violation without requesting permission from users for that new purpose.</p><h4>Without a Posted Notice of Right to Opt-Out, Organizations Must Obtain Affirmative Authorization for the Sale of User Data</h4><p>Additionally, in the CCPA Reasons, there is a rock-solid effort to remove loopholes that could allow a business to sell user data when the business didn’t have a posted notice of right to opt-out, unless the business “obtains the consumer;’s affirmative authorization for the sale.”</p><p>So for all the organizations that didn’t sell user data and didn’t have a posted notice of right to opt-out of data sales, they would be violating CCPA if they turned around and sold the user data without following back up with the consumer for their affirmative consent for the sale.</p><p>The CCPA Reasons from the CA AG also explicitly say why a business can’t just continue to update their data policy notices for new purposes or data sales, because as most people know about user behavior, people don’t go back to revisit privacy policy, terms of service, or data policy webpages, if they even review those pages once.</p><p>From page 15:</p><blockquote><strong>Subsection (e) </strong>was added to state that a business cannot sell personal information it collected during any time it did not have a notice of right to opt-out posted unless it obtains the consumer’s affirmative authorization for the sale. Because Civil Code section 1798.120, subdivision (b), requires a business that sells consumers’ personal information to third parties to provide consumers with notice of their right to opt-out of the sale of their personal information, the converse is also true: if the consumer has not been provided with notice of their right to opt-out when the business collected their personal information, the business cannot sell that consumer’s personal information. <strong>Subsection (e) is necessary to prevent a business from unilaterally and retroactively changing its policy to sell personal information that it collected during a time period when it expressly assured consumers that it did not sell such information.</strong> <strong>If a business decides to change their practice midstream, the business must obtain affirmative consent.</strong></blockquote><blockquote>The OAG considered alternative ways to address this situation and determined that requiring businesses to obtain affirmative authorization is the most effective way to carry out the purpose and intent of the CCPA to give consumers notice and control, at the point of collection, over the sale of their personal information. While the alternative of allowing a subsequently posted notice of right to opt-out to apply retroactively would be less burdensome to businesses, it would not be as effective in informing the consumer of their right at the point of collection, when the consumer may be most aware of what personal information the business is collecting from them. Such an approach would allow businesses to engage in passive notice updates without allowing consumers any agency to control how their personal information is used, including when it was collected under false pretenses. <strong>Furthermore, simply putting up a new notice on a website after a consumer has already provided personal information, when that consumer may be unlikely to revisit the website (and even more unlikely to revisit the notice), is not meaningful consumer notice.</strong></blockquote><h4>CCPA Restricts Service Providers from Retaining or Using Personal Information for Its Own Business Purposes — Direct Attack on COVID Tracking Data Acquired from SDK Companies , Called a “Substantial Modification” by the CA AG</h4><p>There are several clarifications for Service Providers, and there seem to be additional restrictions and clarifications that will apply to any businesses that acquired user data as part of a Service Provider relationship — those businesses are not allowed to retain or use that personal information for its own business purposes.</p><p>Right now, there are a huge amount of analytics companies and mobile app SDK providers that acquired user data as part of Service Provider relationships with other mobile apps — and those organizations have been selling the data for COVID location tracking in violation of CCPA.</p><p>It seems that significant investigations and changes will need to occur based on these sections.</p><p>From page 32:</p><blockquote><strong>Subsection (C ) </strong>has been substantially modified. <strong>The subsection now prohibits a service provider from retaining, using, or disclosing personal information obtained in the course of providing services except to provide those services in compliance with the written contract for services and in four other limited circumstances.</strong> This prohibition is consistent with how the CCPA defines and regulates the disclosure of consumer personal information to service providers and service providers’ use of that information. For the purpose of processing personal information, the CCPA contemplates service providers to be an extension of the business for which it provides services. Civil Code section 1798.140, subdivision (v), defines a “service provider” as one who “processes information on behalf of [the] business” that provided the personal information, pursuant to a contract that prohibits “retaining, using, or disclosing the personal information for a commercial purpose other than providing the services specified in the contract.” Relatedly, a business does not “sell” personal information when it transfers that data to a service provider, provided that the service provider does not “collect, sell, or use the personal information of the consumer except as necessary to perform the business purpose” of the business that provided the personal information. (Civ. Code, § 1798.140, subd. (t)(2)©.) <strong>Thus, the intent of the CCPA is to prohibit a service provider from using personal information collected from one business for its own business purposes or to then provide services on behalf of a different business.</strong></blockquote><blockquote>Many comments objected to the original text of subsection ©, claiming that the CCPA broadly authorizes service providers to retain and use personal information for any “business purpose.” <strong>But nothing in the CCPA allows a service provider to retain or use personal information for its own business purpose. </strong>Rather, as discussed above, services providers are expressly limited from retaining and using personal information. (Civil Code § 1798.140, subds. (v) and (t).) Indeed, the term “business purpose,” when used in the statutory text, contextualizes why a <em>business </em>discloses personal information to a service provider or third party, not the universe of possible ways a service provider could use that information. (See Civ. Code, §§ 1798.115, subd. (a), 1798.130, subd. (a)(4)©.) The CCPA requires that any disclosure of personal information from a business to a service provider be “necessary to perform a business purpose.” (Civ. Code, § 1798.140, subd. (t)(2)©.) Even in defining the term “service provider,” the CCPA makes clear that a business’s disclosure of personal information must be for a business purpose that is stated in the parties’ written contract. (Civ. Code, § 1798.140, subd. (v).) Subsection © thus accurately reflects the CCPA’s requirement that service providers act on behalf of a business by processing information to further the <em>business</em>’s specific business purpose and not for the service provider’s own business purposes.</blockquote><h3>CCPA Innovation &amp; Other Notes from the Final Statement of Reasons from the California Attorney General’s Office</h3><p>The California AG made it clear that the California Data Broker Registry was not only going to be essential for businesses to comply with who are in the business of buying or selling user data, but also pointed out that new industries and privacy innovation can be built with these registries via efforts to standardize global opt-out signals.</p><p>One section on Page 12 included these comments:</p><blockquote>During preliminary rulemaking activities, the OAG learned that a consumer may not know who has and could be selling their personal information, given that the CCPA does not require businesses to disclose the specific persons or entities with whom they shared the consumer’s personal information. (Civ. Code, § 1798.110 [merely requires the disclosure of “categories of third parties” with whom a business shared personal information].) The data broker registry addresses this gap by publicly identifying specific businesses that may be selling the consumer’s personal information. Subsection (e) thus benefits consumers by allowing them to access, in one place, the information they need to exercise the right to opt-out of the sale of personal information from data brokers selling their personal information. It benefits businesses by reinforcing and streamlining their compliance with the data broker registry law and the CCPA. Subsection (e) also encourages the development of consumer tools or services by allowing innovators to pull information about how data brokers process requests to opt-out from a centralized repository.</blockquote><h4>Do Mobile Apps Need to Explain Data Sharing / Selling Before Download?</h4><p><strong>Mobile apps</strong> will be able to include a shorthand reference in their menu and provide links to read more about how the business collects personal information, instead of any required length or specific text.</p><p>But the organizations will still be required to provide details on their data collection on the download pages — basically before a user installs software, details on data collection/sales need to be accessible.</p><p>The reference to a “download page” in these CA AG Reasons could almost be interpreted to require disclosures on App Descriptions before someone installs an app — basically apps need to not only link to privacy policies, but also link to separate pages expressly on how that business collects or sells personal information under the CCPA frameworks.</p><p>From page 13:</p><blockquote><strong>Subsection (b)(1) </strong>has been modified to add that a business that collects personal information through a mobile application may provide a link to the notice within the application, such as through the application’s settings menu. This language was added in response to public comments seeking guidance on whether businesses could include this link through their mobile application’s settings menu. This modification is necessary to clarify that a business has discretion to provide a link directing consumers to the notice in lieu of including the actual language of the notice in the application’s settings menu. It benefits businesses by clarifying requirements for businesses and giving them the flexibility to shorten the language included in the actual application. This modification is not intended to speak to whether a business can provide the notice through its mobile application’s settings menu <em>in lieu </em>of providing it on the application’s download page. In the context of an online service, such as a mobile application, the CCPA defines “homepage” as “the application’s platform page or download page, a link within the application, such as from the application configuration, ‘About,’ ‘Information,’ or settings page, <em>and </em>any other location that allows consumers to review the notice . . . including, but not limited to, before downloading the application.” (Civ. Code, § 1798.140, subd. (l) (emphasis added).) <strong>This requires that the business provide the required information on both the download page <em>and </em>within the application itself, such as through the application’s setting page.</strong></blockquote><h4>Offline customer interactions getting online notices</h4><p>Under the CCPA guidance, businesses that “substantially interacts with consumers offline may satisfy the requirement that it use an offline method to provide notice to consumers by posting signage directing consumers to ‘where the notice <strong>can be found online</strong>.’”</p><p>This full section seems to indicate that at some point in the future, there will be more webpages that are known and that explain how various Point of Sale kiosks, digital billboards and other pedestrian tracking technology is sharing and selling user data. This Reason seems to be another section that will eventually encourage innovation and new privacy products.</p><p>From page 14:</p><blockquote><strong>Subsection (b)(2) </strong>has been modified to clarify that a business that substantially interacts with consumers offline may satisfy the requirement that it use an offline method to provide notice to consumers by posting signage directing consumers to “where the notice can be found online.” This modification was made for the same reasons the change was made in section 999.305, subsection (a)(3)©, above. The modification is necessary to align this provision with section 999.305, subsections (a)(3)©, (b)(3), and (b)(4).</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/627/1*bTyM2xSNBhyejPGTOICK4g.jpeg" /></figure><h4>Everyone hated the “Do Not Sell” Button, AG Agrees</h4><p>In what is potentially the shortest Reason amongst the CCPA guidance from the California Attorney General, there is a short mention of the ill-fated “opt-out button of shame” that was suggested in previous documents from the AG’s office and quickly reduced to ash by UI/UX experts across the internet.</p><p>From page 15:</p><blockquote><strong>Former subsection (f)</strong>, regarding the proposed opt-out button, has been deleted in response to the various comments received during the public comment period. The OAG has removed this subsection in order to further develop and evaluate a uniform opt-out logo or button for use by all businesses to promote consumer awareness of how to easily opt-out of the sale of personal information.</blockquote><h3>Discounts for Sharing / Selling Data, New Ad-Free Subscriptions</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*touBNgYArE7dJshM3_59mA.jpeg" /><figcaption>Hearst Magazines has “<a href="https://optout.hearstmags.com/pubs/HR/A99/AdsFree.jsp?cds_page_id=246801&amp;cds_mag_code=A99&amp;id=15917429885426&amp;lsid=316119844094048392&amp;vid=4">Ad-Free Subscriptions</a>” but this is still a very rare thing to see in the marketplace. Ping me on twitter @ thezedwards if you’ve seen others!</figcaption></figure><p>There are several sections in the CCPA Reasons about providing discounts to consumers for their data.</p><p>I’m not going to excerpt these sections because it’s going to be very hard to thread this needle without violating CCPA, and i’ll need to spend more time on these sections before providing any guidance or opinions about the impacts on various discounting strategies.</p><p>There are some businesses across the United States experimenting with this, and some may have a higher risk tolerance, be simply testing the waters, or be looking to aggressively move into these ad-free markets.</p><p>But one thing is clear, if you’re trying to provide discounts to consumers for ad-free experiences or to sell their data, you should read those sections and consult an Attorney to help you craft the right pricing options and disclosures.</p><h4>Good Clarity: If you don’t sell data, you don’t need to provide a “notice of right to opt-out”</h4><p>Several sections in the CCPA Reasons helped to clarify that businesses would not need to provide a notice of right to opt-out of data sales, if that business doesn’t sell data.</p><p>It’s likely that many businesses complained that this messaging would give consumers the wrong impression about a business, so this section helps to clarify that CCPA is going to be flexible enough in deployment to not provide false signals of data sales.</p><p>From page 21:</p><blockquote>Section 999.306, subsection (d), also provides that a business that does not sell personal information does not need to provide a notice of right to opt-out if it states so in its privacy policy. Thus, the modifications make the language of the regulation consistent with the language in the CCPA and harmonize this subsection with section 999.306, subsection (d). These modifications benefit businesses and consumers by providing clarity and transparency about businesses’ baseline obligations: businesses that state that they sell personal information must post a notice of right to opt-out, <strong>and businesses that do not sell personal information will affirmatively state so.</strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/406/1*crw46f_7QTjZU0qmRjGF3g.jpeg" /></figure><h4>Businesses Urged to Make a “Fact-Specific Determination of What Process to Use” for Consumer Requests to Know/Delete, Only Email Required</h4><p>There are numerous organizations like Mortgage brokers, Banks and Insurance companies that are building complex processes for consumers to safely request to know/access / delete their data.</p><p>For many of these organizations, there were concerns about restrictions being placed on their online forms — but it seems that a business will not be limited by the fields they request from people or authorized agents to complete the submissions.</p><p>The CCPA Reasons also provide some clarity for organizations that operate primarily offline and some assurances to consumers that the primary method they engage with a business needs to have a way to for them to utilize their rights.</p><p><strong>From pages 21–22:</strong></p><blockquote><strong>Subsection (a)</strong>, which governs the methods a business must provide for the submission of consumers’ requests to know, has been modified to provide that businesses operating exclusively online and that have a direct relationship with a consumer from whom it collects personal information <strong>shall only be required to provide an email address for submitting requests to know</strong>. The requirement that businesses operating a website must provide an interactive webform has also been deleted. ..</blockquote><blockquote><strong>Subsection (c )</strong>, which requires a business to consider the methods by which it interacts with consumers when determining which methods to provide for submitting requests to know and requests to delete, has been modified in four ways. First, the word “primarily” has been inserted before “interacts” to clarify the meaning of the subsection. <strong>This clarification is necessary to prevent businesses from designating obscure methods for the submission of consumer requests as a way of discouraging consumers from exercising their rights under the CCPA</strong>. Second, the sentence “At least one method offered shall reflect the manner in which the business primarily interacts with the consumer, even if it requires a business to offer three methods for submitting requests to know” has been deleted. This change is necessary so that the language used in the regulation is consistent with the language used in the CCPA. (See Civ. Code, § 1798.130, subd. (a)(1)(A)-(B).) Third, language has been added requiring businesses that primarily interact with consumers in person to consider providing an in-person method for submitting requests. <strong>The regulation provides a few examples of in-person methods: a printed form the consumer can directly submit or send by mail, a tablet or computer portal that allows the consumer to complete and submit and online form, and a telephone by which the consumer can call the business’s toll-free number</strong>. These examples provide guidance on how businesses should determine which methods to make available to consumers, including those discussed in Civil Code section 1798.130, subdivision (a)(1), while addressing situations in which consumers may need direct, in-person assistance in exercising their CCPA rights. This regulation is necessary to prevent businesses from designating obscure methods for the submission of consumer requests as a way of discouraging consumers from exercising their rights under the CCPA, <strong>while also providing businesses with flexibility to adopt methods that are compatible with their business practices</strong>. …</blockquote><h4>Businesses need to help consumers with CCPA requests, remedy deficiencies in submissions</h4><p>Businesses should create templates for customer support, and are required to provide assistance to consumers who may be unaware of the businesses’ “designated method for submitting CCPA requests.”</p><p>From page 23:</p><blockquote>Consistent with this legislative intent, the regulation provides guidance for instances in which a consumer’s attempt to exercise their CCPA rights is not submitted through a business’s designated methods or is deficient for a reason unrelated to the verification process. <strong>Businesses should provide assistance to consumers who may be unaware of the business’s designated method for submitting CCPA requests or may have made a mistake by contacting the business via some other method.</strong> Furthermore, based on the OAG’s technical expertise in this area and understanding of business practices, treating a consumer’s request as properly received or informing the consumer of the proper method of request is not unduly burdensome. If the business treats a request as properly received, the request proceeds through its designated CCPA-request process. If the business declines to do so, the business can simply provide the consumer with a pre-formulated response with information on how to submit the request and remedy deficiencies.</blockquote><h4>Businesses Should Reply Within 10 Business Days to Requests to Know / Delete, Have 3 Months to Comply</h4><p>A business should test their own processes on a regular basis — if an organization fails to acknowledge receipt of a Request to Know / Delete within 10 business days, or fails to provide additional details within the first 45 day window, or the 45-day optional extension, that business is potentially in violation of CCPA.</p><p>There are several significant sections on the appropriate way to respond to requests, and how quickly these need to be done. The significant details in these sections should remove any doubt that these timing windows are essential for businesses to comply with CCPA.</p><p>From pages 23 and 24:</p><blockquote><strong>Subsection (a) </strong>has been modified in three ways. First, it has been modified to specify that the time period to confirm receipt of a request is 10 “business” days. This change was made in response to public comments seeking clarification of whether the initial proposed “10 days” constituted calendar days and expressing concern that 10 days was not enough time for a business to confirm receipt, particularly when received during business holidays. The clarification of “business days” addresses business holidays and lessens the burden on businesses. Second, the phrase “in general” has been added to clarify that a business’s confirmation of receipt of request simply needs to provide a general description of the business’s verification process. This change was made in response to public comments that requested guidance regarding the level of detail required and that expressed concerns that specific descriptions of a business’s verification process would reveal information to bad actors that could be used to evade security procedures. This modification balances the CCPA’s intent to provide rights and transparency to consumers with the burden on businesses, including potential security concerns. (Civ. Code, §§ 1798.185, subd. (a)(7), 1798.185, subd. (b)(2).) Third, two sentences have been added to clarify that confirmation of the receipt of a request may be made in the same manner in which the request was received, for example by phone. This change was made in response to public comments and is necessary to provide businesses guidance regarding how to confirm receipt of requests. It also reduces the burden on businesses by streamlining the communication methods for receiving and confirming receipt of requests.</blockquote><blockquote><strong>Subsection (b) </strong>has been modified in two ways. First, the word “calendar” has been added to clarify that the time period to respond to requests to know and requests to delete is 45 calendar days. This change was made in response to comments that sought clarification on whether the time period was calendar or business days. Consumers exercising their rights to make requests under the CCPA should not be hindered by unreasonable delays, and 45 calendar days provides businesses with sufficient time to provide the required response, especially considering that they can extend the time to respond by another 45 calendar days. This change is necessary to avoid possible confusion about how to calculate the 45-day requirement. This modification ensures that businesses expediently address consumer requests and prevents excessive wait times for responses.</blockquote><h4>Requests to Know and Delete Must be Verified, Requests to Opt-Out Don’t Require Verification — 15 Day, 90 Day Requirements for Opt-Out Compliance</h4><p>Most organizations provide an “opt-out” through simple immediate mechanisms, but if an organization is working with 3rd parties to sell consumer information, then a series of very important deadlines are triggered when a consumer requests to opt-out of this process.</p><p>The CCPA Reasons includes several details on these deadlines and responsibilities, from page 39:</p><blockquote>Former subsection (e) has been renumbered and combined with <strong>subsection (f)</strong>. <strong>Subsection (f) </strong>now states that a business shall “comply” with a request to opt-out as soon as feasibly possible but no later than 15 “business” days from the date the business receives the request. It also no longer requires a business to notify all third parties to whom it sold the consumer’s personal information within 90 days prior to its receipt of the opt-out request, or to direct those third parties not to sell the consumer’s information. Instead, it requires a business that sells a consumer’s personal information to any third parties after the consumer submits their request but before the business complies with that request to notify those third parties that the consumer has exercised their right to opt-out and to direct those third parties not to sell that consumer’s information.</blockquote><blockquote>The modification that a business comply with the request within 15 <em>business </em>days was made after considering public comments from many businesses and consumer advocates. Some comments called for eliminating the 15-day requirement or extending it to align with the 45-day requirement for responding to requests to know or to delete. Some comments claimed operational difficulties in complying with opt-out requests within 15 days, particularly if requests are received during the holidays, and asked that the regulation at least be modified to 15 business days, not calendar days. Other comments advocated for requiring compliance “immediately” or within 24 hours of receipt of the request due to the immediate nature of the collection and sale of personal information online. The OAG weighed these various comments and determined that 15 business days appropriately balances the right of consumers to opt out at any time with the burden on businesses to process opt-out requests.</blockquote><h3>Significant Clarification of CCPA Definition of Data Sale” Includes “Any Commercial Purpose”</h3><p>In what is potentially one of the more important sections of the CCPA Reasons, the California Attorney General makes it clear that if a business uses consumer data for “any commercial purpose” there will be a “general fairness principle to ensure that a business that is not able or willing to disclose personal information to the consumer cannot profit or commercially benefit from that personal information.”</p><p>This section of the Reasons will need more clarification but I’ve been waiting for some part of the CCPA guidance that could apply to how some businesses upload additional User Data to Google Analytics, and associate that data with UserIDs shared between the business and Google — these match tables are used to improve an understanding of marketing funnels, KPIs, and profitability — and are certainly for a “commercial purpose” and provide valuable new data and context for both the business and Google.</p><p>I’ve tweeted <a href="https://twitter.com/thezedwards/status/1270166350049107968">about this niche issue here</a> and I believe that organizations will need to disclose to users that they are sharing data with Google (and this could apply to other situations of data sharing), and organizations doing this will potentially need to provide the details to users in CCPA Requests to Know.</p><p>From page 26:</p><blockquote><strong>Subsection ©(3)©</strong>, which requires that the business not sell the personal information or use it for any commercial purpose, applies a general fairness principle to ensure that a business that is not able or willing to disclose personal information to the consumer cannot profit or commercially benefit from that personal information. Finally, <strong>subsection ©(3)(d)</strong>, which requires the business to describe to the consumer the categories of records that may contain personal information that it did not search, is necessary to provide transparency to consumers. It informs the consumer that the business may have other personal information about them but assures them that this information is only maintained by the business in an unsearchable or inaccessible format, solely for legal or compliance purposes, and is not being used for the business’s commercial benefit.</blockquote><blockquote>Although some public comments suggested modifying the regulation to permit businesses to fall within this exception if any (rather than all) of these subsections apply, such a modification would too easily allow a business to evade its obligations under the CCPA and could incentivize behavior that would undermine a consumer’s ability to exercise their CCPA rights and access what personal information the business has collected about them.</blockquote><h4>Biometric FaceID Technical Data Not Required to be Disclosed in Requests to Know, Businesses Must Still Acknowledge They Have the Data</h4><p>While not surprising, the CCPA provided guidance that would be useful for companies like Apple and Google that have growing biometric security face scans used to open phones and other devices — those businesses will not be required to disclose this technical data in a response to a request to know, but must acknowledge they have the data.</p><p>From page 26:</p><blockquote>Second, subsection (C )(4) has been modified to add “unique biometric data generated from measurements or technical analysis of human characteristics” to the list of specific pieces of personal information that a business shall not disclose in response to a request to know. This change is necessary to balance a consumer’s right to know with the harms that can result from the unauthorized disclosure of information….Third, subsection (C ) (4) has been modified to require a business to inform consumers with sufficient particularity that it has collected the type of information set forth in the regulation. It also includes a clarifying example. This change is necessary because it provides direction to businesses on what to communicate to consumers when they are prohibited from disclosing these specified pieces of personal information. <strong>It benefits consumers by providing them with information to make privacy decisions while protecting them from the harms that could result from the unauthorized disclosure of this sensitive personal information.</strong></blockquote><h4>There could be a Cold Storage Right to Delete Loophole That Keeps Some Requests in Limbo</h4><p>There’s an important balance between reducing consumer rights and ensuring businesses aren’t overly burdened — the current CCPA guidance seems to provide a loophole for businesses that can’t access an archived or backup system to delete user data. When the archival system is eventually accessed, the business is supposed to then trigger deletion requests that were submitted between the last access point.</p><p>There are potentially scenarios where a business tries to reduce CCPA compliance costs by offloading certain customers (maybe product returns?) into a cold storage location and only accessing it once a year to batch delete any customer requests. It’s probably appropriate to leave this loophole and wait for a business to abuse it, due to this probably being an underused loophole.</p><p>From page 28:</p><blockquote><strong>Subsection (d)(3) </strong>has been modified to allow a business to delay compliance with the consumer’s request to delete only with respect to personal information stored on an archived or backup system until the archived or backup system “relating to that data is restored to an active system or next accessed or used for a sale, disclosure, or commercial purpose.” This change was made in response to comments that were concerned that the initial proposed language “next accessed or used” would be burdensome because the next access or use may be for reasons unrelated to the consumer’s personal information, and that it would deter businesses from implementing reasonable data security practices and procedures because routine maintenance, general testing, or testing of disaster recovery protocols could trigger a deletion obligation. <strong>The comments also contended that many archived or backup systems do not allow specific, targeted deletions, and thus it would not be technically feasible to delete a particular consumer’s information when the archive or backup system was accessed or used. </strong>By modifying the regulation to limit the compliance obligation for deleting personal information on backup systems to when those systems are restored or used for a sale, disclosure, or commercial purpose, <strong>the regulation lessens the burden on businesses</strong>. The modification also preserves the consumer’s right to delete when the business discloses or commercially benefits from access or use.</blockquote><h4>Problematic Privacy Issue: Lists of “Consumers Who Used CCPA Rights” Allowed and Required for 24 Months, Broadly-Defined Suppression Lists</h4><p>One of the worst opinions in the CCPA Reasons, that will potentially lead to very long and invasive forms for Requests to Know/Delete, allows a business to create a global suppression list and “requires a business to maintain records of consumer requests and how the business responded for 24 months” and that the business “may retain a record of the request for the purpose of ensuring that the consumer’s personal information remains deleted from the business’s records.”</p><p>Another way to put this, a business can let you Request to Delete your data, and require that you submit pieces of information to confirm your identity that the business *did not have before you submitted the form* and then the business can hold that information for 24 months. At any point in the future, if the consumer reactivates their account, there doesn’t seem to be an explicit ban on a business merging all customer data, including the data submitted on the Right to Know / Delete forms, into the larger customer account/records.</p><p>I believe the California Attorney General’s office, if they haven‘t already, should clarify to businesses that users should be provided with choice (or businesses flat banned) from merging the data submitted in a Right to Know/Delete into larger customer data profiles, at least without user consent. CCPA should not be used to append new data to customer records, and attempts to do that should only be possible with strong communication to users about that process.</p><p>From page 29:</p><blockquote><strong>Subsection (d)(5) </strong>has been modified in three ways. First, the regulation now correctly cites to “section 999.317, subsection (b),” which requires a business to maintain records of consumer requests and how the business responded for 24 months. Second, subsection (d)(5) has been modified to clarify that the business only needs to inform the consumer of the record-keeping requirement if it complies with the request. Although the record-keeping requirement in section 999.317, subsection (b), applies to all requests received, including those the business denies, <strong>the disclosure to the consumer required here is necessary to dispel any assumption that granting a request to delete will also delete any record of the request.</strong> Notification that the request was denied is unlikely to lead to such an assumption. Third, language has been added to clarify that <strong>a business may retain a record of the request for the purpose of ensuring that the consumer’s personal information remains deleted from the business’s records</strong>. This change was in response to comments seeking guidance on whether businesses can maintain a suppression list. This change benefits businesses by dispelling uncertainty and benefits consumers by <strong>preventing a business from re-collecting information that the consumer had previously requested it to delete</strong>.</blockquote><h4>Service Providers for Public (Government) and Nonprofit Entities Given Disclosure / Deletion Exemptions, Potentially a Loophole for Government Data Brokers</h4><p>This is probably a good but overly broad opinion — it feels like a potential loophole for massive data collection companies that are partnering with governments, and building profiles on ordinary Americans.</p><p>These sections were probably important to include, but these Service Provider exemptions for businesses working with Public and Nonprofit entities will need to parsed, and potentially certain Government Data Brokers not given this same blanket exemption.</p><p>From page 30:</p><blockquote>In light of comments received from the public, the OAG further supplements its statement of reasons in support of subsection (a). (See ISOR, p. 21.) Overall, subsection (a) is necessary to address the unintended consequences that would result from allowing consumers to access and delete personal information held on behalf of public and nonprofit entities and that would otherwise not be subject to the CCPA. Like businesses, public and nonprofit entities outsource operational needs through service providers that essentially perform tasks as if the public or nonprofit entity was doing the task in-house themselves. These public and nonprofit entities also store documents in cloud storage, use email systems provided by third parties, and employ vendors to manage data. For example, a public school district may use a service provider to secure student information, including each student’s grades and disciplinary record. <strong>Without this regulation, service providers used by public and nonprofit entities may be required to disclose or delete records in response to consumer requests because they may constitute businesses that maintain consumers’ personal information.</strong> Service providers for public and nonprofit entities could also be asked to disclose personal information maintained by a government agency, despite the fact that such files may be expressly exempt from disclosure under the Public Records Act.</blockquote><blockquote>Accordingly, the OAG has promulgated this regulation pursuant to its authority to adopt regulations as necessary to further the purposes of the CCPA. (See Civ. Code, § 1798.185, subd. (b)(2).) <strong>Treating a service provider for a public entity or nonprofit as a “business” would not support the purpose and intent of the CCPA because it may expose otherwise exempt personal information to access and deletion requests or force service providers to create unnecessary and burdensome systems to respond to consumer requests.</strong> California law does not provide a right to delete information held by a public entity, nor does it provide a right to access personal information held by a nonprofit entity. The CCPA imposes obligations on “businesses,” which excludes public and nonprofit entities. (See Civ. Code, §§ 1798.100, 1798.105, 1798.110, 1798.115, 1798.120 [imposing obligations on businesses].) It is not intended to allow consumers to know or delete personal information collected by a non-business merely because the non-business outsources tasks to a service provider. In addition, California law already imposes a separate and distinct legal regime to access information held by public entities, including requirements and exceptions that differ from the CCPA. (See, e.g., Gov. Code, § 6250 et seq.) Without this regulation’s clarification, non-businesses, such as public and nonprofit entities, may not be able to employ service providers without risking disclosure or deletion of personal information or without unnecessary and burdensome costs, which may cause them to incur extra expenses to perform operations internally.</blockquote><h4>Opt Out Signals for the Sale of Personal Information Urged to Include Global Signals, New Innovation Needed</h4><p>There is a long history of browsers, publishers, and advertising companies trying to agree on global opt-out signals, and CCPA urges this process to continue and for consensus to be made so that consumers can opt-out via global privacy controls.</p><p>The requirements included in this CCPA guidance are minimal, but the sections are interesting and should be read by anyone working with Consent Management Platforms.</p><p>From pages 36–38:</p><blockquote><strong>Subsection (d)(1) </strong>has been added to provide clear guidance that any privacy control designed or developed should clearly communicate or signal that a consumer intends to opt-out of the sale of personal information. This subsection addresses public comments concerned that a global privacy control may not respect consumer choice, as well as comments seeking clarification on what would constitute a privacy control that communicates the consumer’s choice to opt-out. By requiring that a privacy control be designed to clearly communicate or signal that the consumer intends to opt-out of the sale of personal information, the regulation sets clear parameters for what the control must communicate so as to avoid any ambiguous signals. It does not prescribe a particular mechanism or technology; rather, it is technology-neutral to support innovation in privacy services to facilitate consumers’ exercise of their right to opt-out. The regulation benefits both businesses and innovators who will develop such controls by providing guidance on the parameters of what must be communicated. And because the regulation mandates that the privacy control clearly communicate that the consumer intends to opt-out of the sale of personal information, the consumer’s use of the control is sufficient to demonstrate that they are choosing to exercise their CCPA right.</blockquote><blockquote><strong>Subsection (d)(2) </strong>has been added to clarify how a business must respond when receiving a global privacy control signal for a consumer who has previously agreed to allow the sale of their information, including through participating in a financial incentive program or through a previous business-specific setting. The subsection requires the business to respect the global privacy control signal, but allows the business to notify the consumer of the conflict and ask the consumer to confirm their business-specific privacy setting or participation in the financial incentive program. This subsection is necessary to eliminate confusion by businesses that have received conflicting manifestations of intent from a consumer. It also provides businesses guidance on how to interpret Civil Code section 1798.135, subdivision (a)(5)’s 12-month prohibition on requesting that the consumer authorize the sale of their personal information for consumers who have enabled a global privacy control. Furthermore, this modification benefits consumers by ensuring that they can make discrete choices about the sale of their personal information while still enjoying the ease and reduced friction of not having to submit separate requests to opt-out on multiple websites or applications.</blockquote><blockquote>In light of the comments received from the public, the OAG further supplements its statement of reasons in support of subsection (d) as follows. (See ISOR, p. 24.) Subsection (d) requires a business that collects personal information online to treat user-enabled global privacy controls as a valid request to opt-out. This subsection is forward-looking and intended to encourage innovation and the development of technological solutions to facilitate and govern the submission of requests to opt-out. Given the ease and frequency by which personal information is collected and sold when a consumer visits a website, consumers should have a similarly easy ability to request to opt-out globally. This regulation offers consumers a global choice to opt-out of the sale of personal information, as opposed to going website by website to make individual requests with each business each time they use a new browser or a new device.</blockquote><blockquote>As stated in the ISOR, this subsection is necessary because without it, businesses are likely to reject or ignore tools that empower consumers to effectuate their opt-out right. This is based on the OAG’s expertise in this subject area. As the primary enforcer of the California Online Privacy Protection Act (Bus. &amp; Prof. Code, § 22575 et seq.) (CalOPPA), the OAG has reviewed numerous privacy policies for compliance with CalOPPA, which requires the operator of an online service to disclose, among other things, how it responds to “Do Not Track” signals or other mechanisms that provide consumers the ability to exercise choice regarding the collection of personally identifiable information about their online activities over time and across third-party websites or online services. (Bus. &amp; Prof. Code, § 22757, subd. (b)(5).) The majority of businesses disclose that they do not comply with those signals, meaning that they do not respond to any mechanism that provides consumers with the ability to exercise choice over how their information is collected. Accordingly, the OAG has concluded that businesses will very likely similarly ignore or reject a global privacy control if the regulation permits discretionary compliance. The regulation is thus necessary to prevent businesses from subverting or ignoring consumer tools related to their CCPA rights and, specifically, the exercise of the consumer’s right to opt-out of the sale of personal information.</blockquote><h4>The end, but not all the highlights</h4><p>There are several dozen niche issues throughout CCPA and the Final Statement of Reasons that were not discussed in this already, massive blog post.</p><p>The purpose of this post was to flag some important sections that need to be reviewed by digital strategists, Data Protection Officers, and lawyers working with big data, and flag a few issues that deserve more debate.</p><p>Do you have feedback or think I missed the mark on something? Feel free to respond to the post below or drop me a note on twitter @ <a href="https://twitter.com/thezedwards">thezedwards</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/384/1*SSRlZfZ3Aob85YKNbavQig.gif" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ce63a90d97cc" width="1" height="1" alt=""><hr><p><a href="https://medium.com/golden-data/final-statement-of-reasons-from-the-california-attorney-general-for-ccpa-raises-important-questions-ce63a90d97cc">Final Statement of Reasons from The California Attorney General for CCPA Raises Important Questions</a> was originally published in <a href="https://medium.com/golden-data">Golden Data</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Epic Games Ignored Epic Subdomain Takeover on their Authentication Domain, Promoted $1 Million…]]></title>
            <link>https://medium.com/@thezedwards/epic-games-ignored-epic-subdomain-takeover-on-their-authentication-domain-promoted-1-million-b4d809039b0e?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/b4d809039b0e</guid>
            <category><![CDATA[cyber-security-awareness]]></category>
            <category><![CDATA[php]]></category>
            <category><![CDATA[developer]]></category>
            <category><![CDATA[malware]]></category>
            <category><![CDATA[phishing]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Wed, 20 May 2020 14:51:46 GMT</pubDate>
            <atom:updated>2020-05-20T14:51:46.979Z</atom:updated>
            <content:encoded><![CDATA[<h3>Epic Games Ignored Epic Subdomain Takeover on their Authentication Domain, Promoted $1 Million Bounty to Address User Complaints</h3><h4><strong>A global hacking group took over Epic Games subdomains, then the problem was swept under the rug by Epic Games.</strong></h4><p>At the end of March, 2020, Epic Games posted on their Twitter account a $1 million bounty for anyone to provide information of any corporate astroturfing spreading rumors about Epic Games, particularly with regard to Epic Games’ House Party users complaining about being hacked.</p><h3></h3><p>We are investigating indications that the recent hacking rumors were spread by a paid commercial smear campaign to harm Houseparty. We are offering a $1,000,000 bounty for the first individual to provide proof of such a campaign to bounty@houseparty.com.</p><p>This unusual <a href="https://www.theregister.co.uk/2020/03/31/houseparty_bounty_1m_dollars_commercial_smear_claim/">‘commercial smear’ bounty was covered by a variety of reporters</a>, with a limited amount of facts available, other than random Twitter users complaining about being hacked.</p><p>After the bounty was issued, House Party started to back track, and made a claim <em>rejected by Twitter</em> that there were unauthentic users spreading the rumors. <a href="https://www.buzzfeednews.com/article/katienotopoulos/houseparty-hacking-bounty">Twitter provided a comment to a Buzzfeed article on the story stating</a>, “‘We haven’t seen any coordinated activity related to conversations about Houseparty, but are continuing to keep an eye on it.”</p><p>After the bounty was issued, I quickly just pulled up their web login page to get a sense for their security — I was quickly shocked with their lack of protections for users and tweeted a brief detail about a lack of “Content Security Policy” (CSP) request headers for the login form that would prevent MiiM phishing attacks — their login form could be embedded on a 3rd party website without it breaking:</p><h3>ℨ𝔞𝔠𝔥 𝔈𝔡𝔴𝔞𝔯𝔡𝔰 on Twitter</h3><p>@houseparty Your password reset page @ https://t.co/pE2gLfgvww is on an insecure page that can be embedded for phishing attacks due to no CSP headers // that page is linked from your login page and it&#39;s just one of many problems for your users. This $1 mil bounty = joke, right?</p><p>After tweeting that detail, I realized that there was potentially something much worse going on with Houseparty, and I wrote up a report and tried to submit it to their bounty@houseparty.com email that had been released to the media. Several days later, the email I sent them was responded to, and I was told that in order for House Party to triage my ticket, I would need to submit it through their preferred bug bounty program hosted by HackerOne.</p><p>I re-sent the details of my ticket, and flagged prominently one of the Epic Games House Party subdomains that had been taken over, they looked like this:</p><p><a href="https://hackerone.com/redirect?signature=f619c0c70b6d08f23fc322984a082c6feb30c8b0&amp;url=http%3A%2F%2Fnyc3-prod-worker-138-197-97-151.ms.thehousepartyapp.com%2Flib8%2Fschafzucht.pdf%3Fweb%3Dnyc3-prod-worker-138-197-97-151.ms.thehousepartyapp.com">http://<strong>nyc3-prod-worker-138-197-97-151.ms.</strong><em>thehousepartyapp.com</em>/lib8/schafzucht.pdf?web=nyc3-prod-worker-138-197-97-151.ms.thehousepartyapp.com</a></p><p>There were dozens of similar subdomains compromised by the attackers, and you can still see some of the Redirection Malware that they uploaded via Google searches — this was from May 18, 2020 (the links are broken, but cached content can be viewed):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Z-U43hBIWndCRDfS1UCqKw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RoTqvC4m2wlhVhIrk7Q3SA.jpeg" /></figure><p>If you open any of these cached files that appear to be PDFs in the Google cache, you’ll see random text details and then deep links to tons more of the PDF files across a compromised domain — these deep links push the pages into search results and basically are an SEO-malware-optimization technique. When the pages were live, they also all triggered PHP redirects to new domains — these PDF-&gt;Page URL redirects are visible across the internet on dozens of other compromised domains, more details below.</p><p>Here’s a screen shot of one of the Epic Games compromised pages and the additional requests linking to other compromised content on the Epic Games subdomains — You’ll also see highlighted here a string referencing a unique domain — this string can help you find other compromised websites in this attacker’s network:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hFtIRdVcdM7x_drlNLXxrg.jpeg" /></figure><p>If you take the string above “529591ebd919b17db88db909e28a8” and then Google this string, you’ll find 29 other compromised domains that have this same PDF-string:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KYcU_ZXurioiHtMTY5E9Aw.jpeg" /></figure><p>A similar process can be conducted across all of the malicious PDFs (poisonPDFs) that were uploaded into the compromised Epic Games subdomains — and the attackers even wrote messages to taunt Epic Games like this one in their Lola PDF: <strong>“Rather than enjoying a good book with a cup of tea in the afternoon, instead <em>they juggled with some malicious virus inside their desktop computer</em>.”</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RDXpz6MVRITelZy4WbyjLg.jpeg" /></figure><p>While I haven’t read the book Lola, I’m familiar enough with the content and when it was written to assume that the book didn’t include any references to a “malicious virus inside their desktop computer”</p><p>This string is also useful for finding other compromised domains — a quick search of the initial portion of that string shows that there are even multiple versions of the same sentence, translated slightly differently. You can see those results and more compromised domains:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9nOPvWlvfZKLHM4rF1LvDg.jpeg" /></figure><p>A huge portion of the 44 domains in this search result are still compromised and some of them have lists of all the compromised PDFs — if you click any of these links you will be taken into malicious redirects and I would STRONGLY urge to be extremely careful not only visiting this top level domain but clicking any links like (WARNING: http://www.cloudpeakenergy(.)com/)</p><p>It should also be noted that in the results above, one of the poisonPDFs is for “Roberts Rules of Order” — it’s important to understand that this book is used by politicians, regulators, people in charge of open meetings, and attackers would know that anyone searching for free versions of this book would likely be a VIP individual who could be connected to politicians or political decisions — so attackers like this show their political sophistication by creating entire networks of poisonPDFs for free ebook copies of “Roberts Rules of Order.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JiOtUn5PRoni8IsYqwMbDA.jpeg" /><figcaption>Above, you’ll notice that Miami.edu was amongst the domains that previously hosted poisonPDF malware from this attacker group — their content previously hosted on pages like this Google cached copy @ <a href="http://webcache.googleusercontent.com/search?q=cache:Fn07uKMOxoQJ:info-online.miami.edu/differential-equations-and-linear-algebra-2nd-edition-solutions.pdf+&amp;cd=13&amp;hl=en&amp;ct=clnk&amp;gl=us">http://webcache.googleusercontent.com/search?q=cache:Fn07uKMOxoQJ:info-online.miami.edu/differential-equations-and-linear-algebra-2nd-edition-solutions.pdf+&amp;cd=13&amp;hl=en&amp;ct=clnk&amp;gl=us</a></figcaption></figure><p>Maimi.edu was one of the domains that got hit big time, and they had over 100,000 files hosted on their compromised subdomain before it was fixed:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BaJmasFRJp9O6CKHOuGimg.jpeg" /></figure><p>Whenever you visit one of the hijacked subdomains and their “files” / pages on them, you’ll be pushed into a redirect and dropped into a series of websites that are likely owned by the same criminal group. The domains look like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ssJz3Aid80tVzeMYtTa6GQ.jpeg" /><figcaption>Malicious landing page from @ (WARNING:https://readisthe(.)best/downloads/roberts-rules-or-order-11th-edition)</figcaption></figure><p>If you stay on these pages long enough or click to Register Now, you’ll be taken to a new domain requesting signups — these websites exist for Ebooks, video streaming services, and a wide variety of websites promising “Free Media / Downloads / Books / Movies etc” — here’s one of the landing pages in this specific redirect scam:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Cq1Ka78YpRpFEYtNyrZSbg.jpeg" /><figcaption>Malicious signup from @ (WARNING: https://my-ebooks(.)club/books/signup-spry/#/z=FewwkXjXsa2CjiD25Fqnsh/theme=default/q=Roberts+Rules+Or+Order+11th+Edition/s1=/s2=/s3=/s4=/s5=/source_id=443ed029-2269-503a-ee44-3a36418e5189/project=yYMfRn/mh_offer_id=/dp=36WOjCfD8VJfXttshhErUe/m=/c_bg=/c_img1=/c_img2=/c_color=/source=Referral/software=Browser/domain=cdn.bkc1a(.)club/)</figcaption></figure><p>There are massive networks of these websites — they all have similar messaging, similar promises, and are likely being used for complicated forms of credit card fraud and account hijacks for reused passwords:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*c2syj1jTSeJhKAzmPB_RFg.jpeg" /></figure><p>These websites even signup affiliates and appear to offer fraud services on specific affiliate pages like: <a href="https://my-ebooks.club/affiliate">https://my-ebooks.club/affiliate</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nG8Bm-klzzBCUyD22oDI5A.jpeg" /></figure><p>On this Affiliate signup page, it’s one of the few pages in these attacker networks that actually, factually, is probably used to signup new partners. And even spammers want to know their conversion rates, so the folks who setup this network embedded their actual, factual tracking code from Matomo Tag Manager and several other unique strings can be used to find their other assets. The Matomo Tag string that uncovers more of their work is: https://collecting(.)click/js/container_ecko3JmF.js</p><p>That string -without the (.) included above - can be searched and more of their Ebooks/redirection spam uncovered — some of it captured by antivirus companies:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y-lhQ4Y1IBJBESjTGVZEhQ.jpeg" /></figure><p>If you visit the results in this search, you’ll find more of the “free media” scams like — <a href="https://best-online-books.com/">https://best-online-books.com/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Vk3u1VS40BABk2jKygPG3w.jpeg" /></figure><p>And nearly identical sites can be found using strings on these pages:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OzJpAWB4L_hmBB7IUjdSng.jpeg" /></figure><p>Some of these domains can be searched on SecurityTrails.com and you’ll find subdomains on them that are mapped to larger networks of these hijacked domains — the domains seem to be registered/hosted via “Serverius Holding B.V.” which makes it easy to confirm that the 134 domains registered and hosted on IP Address <a href="https://securitytrails.com/list/ip/37.1.223.152">37.1.223.152</a> on this list are the core domains in their network, which they connect to hijacked subdomains for their redirection scams:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w2TkSqBJ8ruNydg5E65bIQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xYcQUFrh06U5_PcDFeTF8g.jpeg" /></figure><p>These domains all have the same affiliate signups and redirection scams that were highlighted above in the other core domains:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GlKupFg2d3FWGnNuj6P2eQ.jpeg" /></figure><p>On these pages, the web fonts are being requested from “webfonts.ru” — this and other details point to a Russian-language network behind these hijacks:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*czdrkQCtIeoTA4vUONISmQ.jpeg" /></figure><p>These sites are heavily focused on Ebooks, Free movie streaming and other “free media” services that would trick a user into creating accounts quickly — they also have websites like “hd-stream(.)club” which seems to target Soccer fans — if you click on that website to view any of the live streams you’ll be redirected to another site in their network with a URL like @</p><blockquote>https://()<strong>my-sports(.)club</strong>/sports/signup-dual/-/#/z=yhuuIk3ZOQofMgGkUOfQ6Z/theme=default/q=/s1=/s2=/s3=/s4=/s5=/source_id=e7f60728-784f-dd02-117d-e3dc046d292b/project=1M5iou/mh_offer_id=/dp=gTtt4NObPRkknNLoJfIss9/m=/c_bg=/c_img1=/c_img2=/c_color=/source=Direct/software=Browser/domain=<strong>cdn.spba7(.)club</strong>/</blockquote><p>These “Register to Start Live HD Stream” landing pages would likely trick some users into creating simple passwords they’ve used on other websites:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CySy3iwr9wZra07WWYP7EA.jpeg" /></figure><p>The link above in the footer for “Affiliates” — when clicked, takes users to a very familiar Affiliate signup page:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xoy6N-aK0KxkVPGwDZijDw.jpeg" /></figure><p>In the source code of the My-sports.club website, there is a reference to this domain “cdn.spba7.club” — across all these networks in the source code / metadata there are additional domains referenced. A huge portion of these domains, when entered into the URl bar, redirect to “webtools.media” which is a platform SPECIFICALLY built to redirect users to unique webpages/websites based on their country of origin. This is how a user in the United States would be redirected to a specific domain, and someone in the U.K. could be redirected to another domain. This type of orchestration is used by all sophisticated operators to try and hide their work from U.S. auditors or auditors they know exist in specific countries. By merely focusing malicious redirects on certain users and not others, a hijack can look “random” and like some sort of local human error or local compromise, instead of a network compromise orchestrating the delivery of people to specific malware/phishing domains.</p><p>The domains listed @ <a href="https://securitytrails.com/list/ip/37.1.223.152">https://securitytrails.com/list/ip/37.1.223.152</a> have a wide variety of these phishing sites — they all promise high quality content with slick looking websites like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aAz6KjbgxoCQFXqm6-2NrQ.jpeg" /></figure><p>You can view a Google sheet of these domains just in case they go away, <a href="https://docs.google.com/spreadsheets/d/1st5nbUEYwN7nnZkc3UEpvo0XThjr-nc6PNTRqgxPq1c/edit?usp=sharing">here</a>. Be very careful with this list — don’t visit these sites unless you are aware of the risks and purposefully trying to expose yourself to those risks.</p><h3>Criminal Group Running Redirection Scams and Domain/Subdomain Takeovers Hit Epic Games House Party Subdomains, Epic Games Downplays User Risk with Technical Incompetence</h3><p>As mentioned above, a wide range of subdomains were hijacked on “TheHousePartyApp.com” — which is owned by Epic Games.</p><p>The first thing you should notice is that the subdomains that were compromised were on the domain “TheHousePartyApp.com” — the core domain for House Party is “HouseParty.com” but you can see in this screen shot below that the “app.houseparty.com” login page had a Content Security Policy that allowed code to be fired from “wss://*.thehousepartyapp.com” and “https://*.thehousepartyapp.com” // The second entry there with “*.thehousepartyapp.com” has an asterisks “*” in the subdomain field — this is a “catch-all” entry and any subdomain on thehousepartyapp.com then had the elevated privileges to fire javascript code in the window/page of the HouseParty.com login page:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yXRLnjh8hy1H3j7ecLI-qA.jpeg" /></figure><p>The details above, with the CSP policy + the vulnerability that exists on the login page were sent to the Epic Games team, who argued that because the hijacked subdomains didn’t have code from Epic Games on them, somehow those subdomains were safe. They wrote:</p><blockquote>Thanks again for reaching out, all of these records should now be cleaned up and we’re putting checks in place to address future takeovers. Would you like to confirm we haven’t missed any records before we resolve this?</blockquote><blockquote>…</blockquote><blockquote>I’m also dropping this to a Medium, as this was a result of old IPs that a new user inherited, rather than a compromise of our infrastructure. If you have any other questions please let me know! Once you confirm I’ll close this out.</blockquote><p>I asked whether they read my ticket and questioned whether they understood the risk to users, they responded:</p><blockquote>In regards to the severity change. This report was originally submitted under the impression that our environment was compromised, and as a result a bad actor was actively taking over subdomains. <strong>We confirmed that is not the case, and that the subdomains in question were pointing to abandoned DNS records, which in turn were automatically inherited by a third-party which was hosting eBooks</strong>. We found no indication of a targeted compromise.</blockquote><blockquote>We are investigating your findings on api2.thehousepartyapp.com. Can you please file a separate submission so we can track that? Any additional POC you have for that would be greatly appreciated. As long as that submission follows responsible disclosure practices we can transfer it to our bounty program and award the appropriate bounty.</blockquote><p>I further questioned the Epic Games team and their reference to a “Third-party which was hosting ebooks” — they then responded with:</p><blockquote>The domain you listed is a subdomain of thehousepartyapp.com, which is a domain we own. That is one of the subdomains that was addressed in this submission. The content that was hosted there was not being hosted by us. A third party inherited the IP address that HouseParty formerly owned, the DNS record associated with that IP was never removed, which is why that subdomain still directed to the IP in question.</blockquote><blockquote>Even though a subdomain owned by us points to an endpoint, that does not mean the endpoint is hosted by us. For instance, foo.houseparty.com could be directed to Google, but we are not hosting Google’s content. HackerOne has a great writeup on subdomain takeovers available <a href="https://hackerone.com/redirect?signature=ef8bd03b892d933f277325731f19daab9c5b9794&amp;url=https%3A%2F%2Fwww.hackerone.com%2Fblog%2FGuide-Subdomain-Takeovers">here</a>.</blockquote><p>I further questioned their understanding of subdomain takeovers and they replied with more caveats:</p><blockquote>This issue has been discussed with, and investigated by, the Epic Games team and the conclusion is that other than the fact that non-legitimate content could, and was, being hosted by this third party, further exploitation was far less likely.</blockquote><blockquote>You mentioned second order subdomain takeover, and in order for that to happen, domain A, which I assume you are referring to the parent domain, http://thehousepartyapp.com/, would need to import some resources from domain B, the affected subdomain. That is not the case here. The takeover of a subdomain per se does not directly mean that the parent domain would be affected. There are mitigation factors that come into place and render exploitation unlikely.</blockquote><blockquote>In this case specifically, due to the nature of the relationship between the parent domain and the affected subdomain, further exploitation would most likely not be possible. No cookies are set in the affected subdomain and the ability to read/write them in the parent domain would also not make exploitation possible, because session-wise, the parent domain does not rely on session cookies for authentication.</blockquote><blockquote>It is also important to note that when dealing with security vulnerabilities in the bug bounty realm, evidence of exploitation along with a proof of concept leveraging the issue to a practical exploitation scenario are of utmost importance. Even though you have pointed out that some subdomains pointers were left dangling and a 3rd party was hosting content on those assets, everything else is purely theoretical.</blockquote><blockquote>All of the above results in the reasoning behind lowering the severity of the issue, and why it did not have as severe a potential impact to Epic Games as it might have originally seemed. Hopefully that will help to explain the situation, and how the team came to their conclusion.</blockquote><p>… and also wrote:</p><blockquote>The first discussion is related to the technical issues that you submitted. Those being, that there were a number of subdomain pointers that had not been updated and were pointing to an IP address now controlled by another organization, and the permissive cookie structure in the CSP. This conversation is relevant to your submission to the HackerOne platform and, with the information that you have provided, the Epic Games team evaluated the impact and removed the abandoned DNS entries identified in your report. The team also identified and removed additional entries no longer in use on the HouseParty platform. Epic Games is implementing further tooling to address retired subdomains and evaluating updates to the CSP on app.houseparty.com. To be clear, these subdomains were not maliciously created by an attacker, they were created by the HouseParty team and were not deprovisioned when they were replaced. The Epic Games team has found no indication of exploitation of the DNS entries you identified, nor any additional entries identified by the team.</blockquote><blockquote>The standard flow of how reporting works on the HackerOne platform is that a hacker provides a proof of concept for a security issue and how it could be reproduced. Then the security team at an organization internally reviews the impact of the finding and takes any appropriate actions. In this case, any security vulnerabilities related to your finding are theoretical from your point of view, of what <em>could</em> be done, not what has been proven. There is a lot of discussion in this report so, if you are able, please provide a new technical write up of any proof of wrongdoing that you were able to identify, and we will be happy to take that into consideration.</blockquote><blockquote>As a side note, the issue that you mentioned relating to api2.thehousepartyapp.com sounds like a separate issue, and the Epic Games team would encourage you to submit that separately to the program on HackerOne. They will be glad to evaluate it separately, please keep in mind this submission must follow responsible disclosure in order to qualify for a bounty on HackerOne.</blockquote><h3>Epic Games was even told the name of the attackers, a link to a news article about the attackers, and details about these phishing / credit card attacks that have gone on for years</h3><p>I’ve named the attackers who commit this work the “Pickaflick.com Crew” due to the fact that they have been operating various credit card scams for a decade or more (and were basically named this in the past) — you can see their legacy “PickaFlick.com” website @ <a href="http://web.archive.org/web/20140121175615/http://pickaflick.com/">http://web.archive.org/web/20140121175615/http://pickaflick.com/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xSvrtAZCU4rC-ZBqKov8ag.jpeg" /></figure><p>This organization can be reverse engineered using some of the strings in this research and many others — it must be understood that this entire architecture is probably built by just a few super organized people who uses PHP orchestration to automate their fraud — this automation leaks shared strings across environments. You can find strings in the PDFs and easily find other compromised domains like this search that has 8,440 results: <br><a href="https://www.google.com/search?q=%2258831883288308839%22">https://www.google.com/search?q=%2258831883288308839%22</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ADtKsysMGPknQgFibOk1dA.jpeg" /><figcaption>Some of these domains have fixed their redirection SPAM from the PickaFlick crew, but many haven’t…</figcaption></figure><h3>Architecture of Credit Card and Phishing Fraud</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B89E2yDEc9lB2e3PEU9C7A.jpeg" /></figure><p>It’s quite easy to find the compromised domains in this network and the PDF files they use to orchestrate the PHP URL redirects — you’ll find lots of domains like <a href="https://movie-frame.com/">https://movie-frame.com/</a></p><p>On several of these domains, in the footer, a corporation and address in Cyprus is referenced: <strong>Varvelia Technologies Ltd Prodromou, 75 1st floor, Flat/Office 101 Strovolos, 2063 Nicosia Cyprus</strong></p><p>Here are the <a href="https://opencorporates.com/companies/cy/HE346627">OpenCorporates records for Varvelia Technologies</a> — I’m not going to include any other details or potential research on this organization because it seems likely that a real law firm potentially helps one of their clients register these domains, but you can see that many of their domains are built with nearly identical messaging, stolen content, fake offers, and they reference Varvelia Technologies in the footer:</p><blockquote><a href="https://t.co/LMfwoOUjgi?amp=1">https://silverscreen.cc</a></blockquote><blockquote><a href="https://t.co/4exotBALBp?amp=1">https://crazyebooks.com</a></blockquote><blockquote><a href="https://t.co/o0cn2NLV9M?amp=1">https://cinefique.com</a></blockquote><blockquote><a href="https://t.co/vPfBdjzEUd?amp=1">https://entertainment-lab.com</a></blockquote><blockquote><a href="https://t.co/CtWvLBOUwW?amp=1">https://movie-river.com</a></blockquote><blockquote><a href="https://t.co/QWa2Hjhv0u?amp=1">https://river-support.com</a></blockquote><blockquote><a href="https://t.co/8ektgL7Muk?amp=1">https://ceb-support.com</a></blockquote><blockquote><a href="https://t.co/e6qycsYW97?amp=1">https://support-frame.com</a></blockquote><blockquote><a href="https://t.co/SXDJgXb3MX?amp=1">https://mrv-info.co</a></blockquote><blockquote><a href="https://t.co/tYuTmmPxMl?amp=1">https://screen-support.com</a></blockquote><blockquote><a href="https://t.co/zKkU6VDnUg?amp=1">https://paf-user.co</a></blockquote><blockquote><a href="https://t.co/lVhaWqdelw?amp=1">https://mfr-aid.co</a></blockquote><blockquote><a href="https://t.co/EQ0u1KmYFs?amp=1">https://q-help.org</a></blockquote><blockquote><a href="https://t.co/Xh19hFyevJ?amp=1">https://slv-aid.co</a></blockquote><blockquote><a href="https://t.co/Trgn0KevJy?amp=1">https://ebookfrog.com</a></blockquote><blockquote><a href="https://t.co/1HZjxWelhX?amp=1">https://movie-frame.com</a></blockquote><h4>Custom PHP malware for special domains from PickaFlick crew</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*duIIkgBkmi5nSGJiRzuo9w.jpeg" /></figure><p>Some of the domains above had users from very large domains being pushed into them, and due to that, some of those requests were fortunately captured by URLscan several months ago — this helps to prove that the group has custom PHP malware that they deliver to certain domains.</p><p>The University of Toronto is a good example of this — they had their “Giving.Utoronto.ca” subdomain compromised for a significant period of time and thousands of PDFs were uploaded and cached into Google search for users to find:</p><p>100% of the URls on the Giving subdomain are now broken (but cached results are easy to find) — but due to how much traffic was pushed through these subdomains, at some point, you can find <a href="https://urlscan.io/result/e33a0214-c9c7-468b-bedb-9b7f12cbca85/#transactions">cached versions of these pages in URLscan.io </a>and those requests show that this PHP fraud network had the ability to inject a fake PHP captcha into pages and the URL redirects to compromised Russian Yandex analytics packages to track the opens:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Bic9hd5vUhTBvYRbMCWhQg.jpeg" /></figure><h3>PickaFlick.com Complaints in BBB, News Coverage Prove Multi-Year Campaigns</h3><p><a href="https://www.bbb.org/us/ny/new-york/profile/internet-service/pickaflick-0121-169440/complaints">The Better Business Bureau has a webpage for PickaFlick</a> that includes complaints from 2017 and 2019 with users complaining about very small fraudulent credit card charges and the company requesting copies of credit card and debit card statements to get charges removed:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xRGOGOie8CjX5cKRCt877A.jpeg" /></figure><p>In May 2017, <a href="https://www.wmcactionnews5.com/story/35323566/unauthorized-credit-card-charges-showing-up-on-movie-streaming-service/">WMC5 Action News and ScamAdvisor.com did an investigation into PickaFlick.co</a> (the group originally used ‘PickaFlick.com” but changed ~5 years ago according to the Way Back Machine) — and the investigators found them to be scammers — and they raised very serious questions about whether these networks were being used to submit fake credit cards, and by pushing real users into these networks the stolen credit cards and transactions could be hidden amongst real users who accidentally were signing themselves up.</p><p>A few details from the article linked above:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U7Dx3T8SA0RWTjHLVYZNhA.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3W5o9BSAp7-Y7XRECqKQog.jpeg" /></figure><p>This same scam was highlighted by OnlineThreatAlerts.com in 2017 where they called it “<a href="https://www.onlinethreatalerts.com/article/2017/7/16/beware-of-www-pickaflick-co-it-is-a-fraudulent-online-streaming-website/">A Fraudulent Online Streaming Website</a>” and went on to connect the work to several similar networks:</p><blockquote>“www.pickaflick.co&quot; is a fraudulent content streaming website that claims online users can watch their massive library of movies &amp; TV Shows in glorious HD. Pickaflick is similar to the fraudulent websites: <a href="http://www.geeker.com,/">www.geeker.com,</a> lilplay.com, and tzarmedia.com. Therefore, online users who have used their credit cards on the fraudulent website (Pickaflick) run the risk of their cards being fraudulently charged for undetermined amounts, even when they have canceled their subscriptions”</blockquote><h3>Were the Epic Games House Party Users Hijacked? Ubiquiti Inc Hackerone Ticket shows risk of subdomain auth takeovers.</h3><p>As part of my efforts to explain to Epic Games the risks of an auth subdomain takeover by the PickaFlick crew, <a href="https://hackerone.com/reports/172137">I shared a public ticket showing how a security researcher hijacked an Ubiquiti subdomain</a> and explained how it could be used to attack users…</p><p>I’d urge reading the entire ticket because this is the risk that organizations face when they have subdomains compromised by serious organizations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kLXPOEkXzMwqVHfPKJP9eA.jpeg" /></figure><p>This information above was largely disregarded by Epic Games, and instead of considering the risk to their users, I was asked for “Proof of Concept” that the subdomain takeovers could lead to auth takeovers or other attacks on Epic Games users — I largely told Epic Games that they were being ridiculous to ask for a POC when I was pointing out hijacked subdomains — there was no “concept” involved — it was a real attack, with real subdomains on the Epic Games auth domain hijacked, at the time of my ticket submissions. Epic Games even removed the subdomains, but continued to downplay the risk to their users, and they refused to consider who this network was, what other organizations they’ve hit, their obvious technical sophistication, and their likely connections to credit card fraud and other types of user phishing fraud.</p><h3>Conclusion:</h3><p>Credit card fraud is complicated, so is phishing user credentials through authentication-based-session attacks and PHP redirects.</p><p>The process to hijack subdomains, even when most appear to be using DigitalOcean (the excuse Epic Games provided basically), is not insignificant, and this group obviously has massive resources, even though some coding practices they need to reuse in order to scale their systems (and therefore provides the strings to find their networks).</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FQ8gTu3iksU65Y3Mn8R%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fschittscreek-schitts-creek-pop-tv-poptv-Q8gTu3iksU65Y3Mn8R&amp;image=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2FQ8gTu3iksU65Y3Mn8R%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="435" frameborder="0" scrolling="no"><a href="https://medium.com/media/67d332347bb41cad2df7d0a4ac6f5a06/href">https://medium.com/media/67d332347bb41cad2df7d0a4ac6f5a06/href</a></iframe><p>I believe the facts are clear that an organized hacker and credit card fraud network used Epic Games subdomains to launch attacks on users. And this group continues to orchestrate these attacks across the internet, and has for years.</p><p>Finally, if you are going to offer a $1 million bounty to find out who could be attacking your users, when confronted with this research, how could you reject all the conclusions that this was the organized group targeting you and your users?</p><p>Have problems, concerns or would like to see edits here? Ping me @ thezedwards on Twitter and i’ll see what I can do. Thank you! 🖖</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b4d809039b0e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The 2020 URL Querystring Data Leaks — Millions of User Emails Leaking from Popular Websites to…]]></title>
            <link>https://medium.com/@thezedwards/the-2020-url-querystring-data-leaks-millions-of-user-emails-leaking-from-popular-websites-to-39a09d2303d2?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/39a09d2303d2</guid>
            <category><![CDATA[gdpr]]></category>
            <category><![CDATA[advertising]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[data-leak]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Wed, 29 Apr 2020 15:49:34 GMT</pubDate>
            <atom:updated>2020-04-29T15:49:34.619Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2aUd6Kp_TbD4wTgVqqV1Tg.jpeg" /></figure><h3><strong>The 2020 URL Querystring Data Leaks </strong>— Millions of User Emails Leaking from Popular Websites to Advertising &amp; Analytics Companies</h3><h4>Breaches have been found on websites including Wish.com, JetBlue.com, Quibi.com, WashingtonPost.com, NGPVan.com and numerous other organizations…</h4><p><strong>Most popular websites on the internet are using 3rd party analytics and advertising Javascript code</strong> — and depending on how a website sets up their marketing systems, typically email systems and new user signup flows, the user emails can accidentally and/or purposefully leak to companies across the global data supply chain.</p><p>The organizations included in this research have hundreds of millions of emails and real users between them — and only Wish.com, Mailchimp and The Washington Post took this report on their user email breaches seriously — Wish updated their email system within ~72 hours of the report being sent and the other two started taking actions relatively quickly — whereas many other organizations either didn’t respond or have failed to take any actions for weeks or months.</p><p>All organizations need to be aware of this significant user data vulnerability, but more importantly, <strong>there needs to be significant efforts by organizations sharing user emails in this way, to submit *partner deletion requests* to the 3rd party advertising and analytics companies who received the user emails.</strong></p><p>Throughout this research, some of the advertising companies that were tracked receiving the user emails are included — but this should not be considered 100% complete due to the fact that this has been going on for years on some websites, and it’s impossible to know externally all of the organizations who synced data on a specific website or webpage at any historical point.</p><blockquote><strong>All organizations included in this research leaking user emails should publicly post the list of all their historical advertising and analytics vendors who could have received the user emails while their respective breaches were active.</strong></blockquote><p>One important trend to notice is how often Google Analytics, Google’s DoubleClick, Facebook, and Twitter are ingesting the user emails — these are organizations that should be receiving deletion requests en-masse and they should all have processes to handle this type of effort already (Facebook likely has this tech already based on conversations on this research and additional research from a private report from several years ago).</p><p>In this research, there are also “red flag organizations” who have ingested user emails that are small or relatively unknown organizations, yet likely receiving huge amounts of user emails in their request logs. These smaller organizations need a unique type of scrutiny due to the power that an advertising or analytics company can attain from ingesting millions of user emails from their enterprise clients — <a href="https://www.theguardian.com/uk-news/2018/may/06/cambridge-analytica-kept-facebook-data-models-through-us-election">the Cambridge Analytica effect if you will</a>…</p><h3>Numerous Enterprise Organizations Leaking User Emails Through 3rd Party Javascript Request Headers Sent via Browsers to 3rd Party Advertising &amp; Analytics Companies</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7oWacfG-yGSyqZDVUL3lbQ.jpeg" /><figcaption>Each of these orgs leaked user emails by unsafely appending the user email to a URL in plain text (or encoded in base64) and then having the user emails leak to 3rd party advertising and analytics companies.</figcaption></figure><p>When any 3rd party Javascript code loads on a website, metadata from the user and the website can be transmitted to the 3rd party domain / company that controls that code — this is technically through the “Request Headers” sent through a browser — and this data can include what page a user is visiting, what type of device and browser they are using, their location, and other forms of fingerprinting / cookies / URL querystring/ URL parameters that are used by advertising and analytics companies.</p><p>This type of email user data in a URL bar synced into Javascript pixels is most typically blocked by a regular person through “Ad blockers” or through browsers like Safari, Brave, and Firefox — those browsers use Javascript/cookie blocking as a default features to protect users (each browser handles it slightly differently). This breach and research included here would impact all Chrome users of these websites who went through these specific user flows and who didn’t proactively block all Javascript (a rarely used option) or use a Chrome “Ad blocker” extension that blocked this type of Javascript. Some people using the other “safe” browsers (Safari/Brave/Firefox) could have been protected from the leak due to their 3rd party Javascript requests being blocked.</p><p><strong>Most of the data breaches that were found (some are still live breaches as of publishing) are caused by a sloppy and dangerous growth hack that is used to improve attribution tracking for analytics tools and used to optimize and segment retargeting advertising campaigns.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bFZJ8wcbZLEiw7Q_p1AWfg.jpeg" /></figure><p>Several of the breaches involve “plain text” user emails — this is when you can literally read the email in the URL with minimal changes/encodings.</p><p>Some of the breaches involve a form of plain text known as “base64 encoding” — in short, base64 is a programming language feature that is NOT a form of encryption and provides no user protections. A base64 string can be decoded through many tools, and there is even a free service from the s̶p̶i̶e̶s̶ nice folks at GCHQ called <a href="https://gchq.github.io/CyberChef/&#39;">CyberChef</a> for parsing custom base64 encodings.</p><p>Before I get into the details about how this breach happens, and the specific circumstances surrounding the examples, I want to briefly acknowledge and give credit to the team at Wish.com for how quickly they changed their entire email architecture after being informed of their breach — in less than 72 hours Wish had completely rebuilt their email architecture and they had built a completely new auto-login flow via email.</p><p>I believe the Wish.com breach was the largest out of all the examples in this research, and it lasted over a year and likely involved hundreds of millions of user emails in a base64 plain-text format being shared with analytics and advertising companies, but their work to quickly escalate the problem, realize the scope, and then pull the trigger to rebuild their systems was a dramatically better response than how other organizations handled these reports. I believe Wish and all organizations in this research should be requesting deletion of user emails from any 3rd party logs held by external advertising and analytics companies, but it appears no organization has submitted this request to their partners, even after being notified of their breaches.</p><p>For the most part, most of these user email data breaches are <strong>still live</strong> as of publishing this research — and in this research I’ll show you how to “breach yourself” by just using current website signup flows and other normal website features on the specific websites in question.</p><p>I also want to thank <a href="https://twitter.com/eliyastein">Eliya Stein</a> at Confiant.com for being a sounding board on these technical issues, and helping to provide an additional vet and other important context around the Wish.com breach (those details below).</p><h3><strong>3rd Party Javascript Collects a “</strong>Referrer<strong>” URL Field, Which Can Leak User Data and Email Addresses from a Website</strong></h3><p>This research is focused on a specific type of user data breach that occurs due to how Javascript collects data on a website. When a user loads a web page, the URL that they are visiting, along with any URL parameters (extra tracking codes appended after a “?” in a URL) are shared with any advertising or analytics companies through the javascript code on that page and through a technical browser transmission “request header” known as a “Referrer” field.</p><h3>Quibi Leaking New User Emails on Email Confirmation Webpage to Advertising and Analytics Companies</h3><p><em>(Pre-Publishing Note: Quibi reached out hours before publication with an apology and several sentences explaining “how this happened” and what they were doing to fix it. Apparently they no longer leak user emails — I have doubts about some of their statements and will let other reporters publish their remarks)</em></p><p>When you install the Quibi app, you are asked to submit an email to create your account, and then emailed a confirmation link that must be clicked to confirm the account. When a user clicks this email confirmation link, their email address is appended into the URL they are clicking in plain text, and sent to 3rd party advertising and analytics companies.</p><p>Quibi was informed of their user email data breach on April 17, 2020 but haven’t responded to the details other than through their automated customer support system.</p><p>Here’s a screen shot showing the Quibi New User Email Verification Webpage URLs and how this page was built to leak the user email in plain text to advertising and analytics companies:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IgYe_w1UMMIAPZnw-05ouw.jpeg" /></figure><p>That same “Email verification” webpage above from Quibi sends the data to advertising and analytics companies through the referrer fields in the request headers — a screen shot below includes the user email sync from Quibi to Snapchat’s <em>sc-static.net</em> advertising endpoint.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y1ZINUYXsNIWtL83pWIxrg.jpeg" /></figure><p>Here’s a screen shot of the Twitter request as it receives the user email in the URL:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Vyo48REkxyKA6ZZTq0iCBw.jpeg" /></figure><p>Here’s what one of the email confirmation links looks like:</p><p><a href="https://quibi.com/email_verified/?email=quibi%40victorymedium.com&amp;message=This%20URL%20can%20be%20used%20only%20once&amp;success=false&amp;_branch_match_id=759077528166021115&amp;utm_source=Email&amp;utm_campaign=Account%20Management&amp;utm_medium=Email%20Verification#">https://quibi.com/email_verified/?<strong>email=quibi%40victorymedium.com</strong>&amp;message=This%20URL%20can%20be%20used%20only%20once&amp;success=false&amp;_branch_match_id=759077528166021115&amp;utm_source=Email&amp;utm_campaign=Account%20Management&amp;utm_medium=Email%20Verification#</a></p><p>When initially tested, the user email address in plain text format was transmitted to:</p><blockquote>1) Google’s DoubleClick.net endpoint</blockquote><blockquote>2) Google’s updated ads endpoint @ <a href="http://google.com/">google.com</a></blockquote><blockquote>3) Google Tag Manager (and therefore potentially custom tags could fire for specific visitors/geos/URL params, thus leaking this to more companies)</blockquote><blockquote>4) Twitter ads endpoint</blockquote><blockquote>5) Snapchat ads endpoint &amp; the <a href="http://tr.snapchat.com/">tr.Snapchat.com</a> subdomain</blockquote><blockquote>6) Google Cloud infrastructure via <a href="http://cloudfunctions.net/">cloudfunctions.net</a></blockquote><blockquote>7) CivicComputing.com, which redirects to <a href="https://www.civicuk.com/">https://www.civicuk.com/</a> and appears to be a company based in the United Kingdom.. this raises big GDPR red flags….<br>8) Facebook events / custom audiences for ads</blockquote><blockquote>9) Google ads conversion pixel</blockquote><blockquote>10) Twitter ads conversion pixel</blockquote><blockquote>11) Google Analytics</blockquote><blockquote>12) Facebook analytics, Google Analytics, Twitter analytics (they fire at the end of the page load again)</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Wc6mKG9hjBYQ9EEZpaKNJw.jpeg" /></figure><p>The Quibi new account email confirmation flow was tested again on April 26, 2020 and it was confirmed that the user email is still being appended to the email confirmation page URL in plain text and leaked to 3rd party advertising and analytics companies.</p><p>Since the original test, several new advertising companies were found receiving the user data including LiveRamp.com, SkimAds, and Tapad — it seems likely that numerous ad tech orgs have been syncing the Quibi new user emails and the list included here could be incomplete.</p><blockquote>Quibi’s user data breach is one of the most egregious in this research, because they are a new and extremely well-funded organization and were launched well after both GDPR and CCPA went into effect. In 2020, no new technology organizations should be launching that leaks all new user-confirmed emails to advertising and analytics companies — yet that’s what Quibi apparently decided to do.</blockquote><p>Out of all the data breaches in this research, the Quibi research is the hardest to swallow due to how new this organization is, and how much money they had to push into their marketing and advertising to grow new users — it’s an extremely disrespectful decision to purposefully leak all new user emails to your advertising partners, and there’s almost no way that numerous people at Quibi were not only aware of this plan, but helped to architect this user data breach.</p><p>It’s 2020, and this type of growth-hack needs to stop being green lit. Quibi needs to explain to their users why this was done and why it hasn’t been changed even after being notified…</p><h3>The Biggest Breach: Wish.com Likely Leaked Hundreds of Millions of User Emails for Over a Year, With the User Emails Encoded into Base64 Strings</h3><p>From July 2018 until January 2020 when this research was initially shared with Wish.com, Wish transmitted user emails to at least Google, Facebook, Pinterest, Criteo, PayPal and Stripe, and potentially other companies.</p><p>In July 2018, Wish.com deployed code that started their user email breaches — this was tracked due to user emails in base64 format being cached in systems like URLscan.io — the Wish.com developers deployed code that started to encode users emails in base64 plain text and then append that string into URLs sent to users via email in a URL parameter named “ee” — when users clicked on any marketing emails from Wish, their email was appended to the URL for any page/product-page they clicked from the marketing emails from Wish, and then when the user visited the Wish page, their email in base64 format was transmitted to Wish’s 3rd party advertising and analytics partners.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_AI4bmzaifhAFFco7f7K4g.jpeg" /><figcaption>A URLScan.io capture of a Wish.com page view from July 23, 2018 that captured a user’s “ee” parameter and their email encoded in base64 plain text. The ee string is blurred for privacy due to it containing the user email.</figcaption></figure><p>Approximately ~72 hours after being informed of this research Wish rebuilt their entire email architecture and stopped appending the “ee” parameter with base64 user encoded emails into marketing emails. It does not appear Wish has informed their users of this user email breach, but they did take the issue very seriously and quickly agreed that the base64 email encoding was a practice they weren’t going to continue. Minimal comments from Wish were received after the research was submitted, but they were more professional than the most organizations when confronted with this type of research.</p><p>Due to Wish.com being a massive multi-billion dollar company, <a href="https://www.vox.com/2015/12/28/11621724/meet-wish-the-3-billion-app-that-could-be-the-next-walmart">who in 2015 was Facebook and Instagram’s #1 app advertiser over Christmas, spending upwards of $100 million</a>, and their previous valuations, it’s likely that tens of millions, if not hundreds of millions of user emails were pushed through the “ee” parameter and leaked to advertising and analytics companies.</p><p><strong>To repeat: from 2018–2020, most if not all of the Wish.com marketing emails appended user emails in a format that, if the user clicked on the email and they were using a browser that didn’t block 3rd party javascript, then that user had their email in base64 plain text format leaked to 3rd party advertising and analytics companies including Google, Facebook, Pinterest, Criteo, PayPal and Stripe, and potentially other companies.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*4YaSe1kVP-iAdMAsQrhCJw.png" /></figure><p>The URLs being shared by Wish during this period looked like this (my base64 email is replaced below with XXXXX):</p><p>https://www.wish.com/feed/xparam-5e1ca48aac2ad7067968f60b?utm_campaign=5e1ca314ac2ad7067968f60a&amp;uuid=214e5a2c231841ca89c7bb953681de36&amp;cmpgnid=5e1ca314ac2ad7067968f60a&amp;<strong>ee=XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX</strong>&amp;email_section=core_cids_0&amp;rerank=5a6c31baff96b67f61800650&amp;exzpl=ctp-1&amp;filter=xparam-5e1ca48aac2ad7067968f60b&amp;utm_medium=email&amp;utm_source=Wish+Discount&amp;recvuid=5de37c007ad4122b6cdd8bc6&amp;iscommerc=1</p><p>At least thousands of Wish.com users had their base64 email address cached into Google search results, URLScan and other public systems — today, you can still search for this on Google, and a huge portion of the results, you will see are actually the “ee={user-base64-emails}” string that leaks via email clicks: <a href="https://www.google.com/search?q=site%3Awish.com+inurl%3Aee">https://www.google.com/search?q=site%3Awish.com+inurl%3Aee</a></p><h4>Confirmation from Eliya Stein at Confliant.com, Including Flagging Several New Organizations Receiving Data from Wish.com</h4><p>Early in this research, Eliya Stein (<a href="https://twitter.com/eliyastein">Twitter</a>, <a href="https://www.linkedin.com/in/eliya-stein-74b9007/">Linkedin</a>, <a href="https://www.confiant.com/">Confiant</a>) was contacted for a quick technical double check, and helped to identify several new service providers that had been receiving data from the Wish.com base64 user data breach. Eliya’s concise report is included here:</p><blockquote>It didn’t take long for Wish to send me a marketing email and I was able to confirm the finding immediately.</blockquote><blockquote>When a subscriber clicks a link in the email, the destination URL has several parameters in the querystring, including “ee” which is paired with the recipients base64 encoded email address.</blockquote><blockquote>This means that this entire URL, including the ee parameter can potentially be leaked to any 3rd party resources that are loaded on the page.</blockquote><blockquote>In this case, I can confirm that it’s being leaked at least to Facebook, Google, Pinterest, and Critieo per Zach’s observation, but also additional 3rd parties including Paypal &amp; Stripe.</blockquote><blockquote>From my observations, it looks like these are mostly tracking endpoints and not actual ad slots on these pages. If they ever introduce display ads connected to rtb on these pages, then the impact of this leak has the potential to be quite large.</blockquote><blockquote>I’m not in a great position to comment on GDPR implications, because that’s a little bit outside of my expertise, but for sure it’s a terribly bad practice to pass around PII in plain text in the URL like this, and I do consider base64 encoding to be plain text.</blockquote><blockquote>One thing that we’re not able to observe is if and how this data is being abused, but any ad tech company with integrity should scrub data like this if they recognize it as PII.</blockquote><blockquote>I’ve included a screenshot of the parameter being leaked to Facebook via the referer.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LjBAUIp3KpRHemEdKwXfmg.jpeg" /></figure><p>Wish.com, like the other organizations included in this research, would ideally submit deletion requests to all of their advertising and analytics partners who received data during this period with requests to those partners to delete the request logs containing base64 user emails.</p><p>Ideally, Wish.com would also inform their own users about this breach too.</p><h3>JetBlue.com Still Leaking New User Emails to Advertising and Analytics Partners</h3><p>Jetblue has known about their ongoing data breach since March 2020 and sent several email responses after being shown this research, but still haven’t made any changes to their website or the ongoing leak of new user emails to 3rd party advertising and analytics companies.</p><p>After being informed of the leak, Jetblue stated they would never do what they are doing because it would be against the law (<em>*NOTE: JetBlue wrote “Federal Passenger Privacy Act” in their response — this may be a reference to a 1974&#39; privacy bill — or as </em><a href="https://www.law.berkeley.edu/files/sb27report.pdf"><em>this Berkley Law paper on page 14 indicates</em></a><em>, JetBlue has sent this statement before and is possibly referencing a nonexistent law</em>), writing this in a response recently:</p><blockquote>We regret to hear of any disappointment you experienced when creating a TrueBlue account. We can assure you we don’t share your information. <strong>The Federal Passenger Privacy Act* strictly prohibits the release of any information regarding our customers</strong> or their travel to any other party. We even require specific security information to verify the identity of our customers before we’re able to discuss their own information.</blockquote><p>These details were tweeted out in March and then emailed to JetBlue, which still didn’t have an impact to get them to change:</p><h3>ℨ𝔞𝔠𝔥 𝔈𝔡𝔴𝔞𝔯𝔡𝔰 on Twitter</h3><p>Howdy Data Supply Researchers, tired of sitting on all this research I have so i&#39;ll drip some out right now. JetBlue claims they don&#39;t have an active data breach today, that they don&#39;t share the user email on registration w/ ad tech companies. Test me https://t.co/4aKbLxCRX8</p><p>Here’s the flow of how all new JetBlue users are having their email addresses leaked to 3rd party advertising and analytics companies, in violation of the Federal Passenger Privacy Act* (and potentially other privacy laws)— step one, click “Join” in the menu bar on Jetblue.com from the homepage or any page on the site:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tWel2qV81CyRFH-L1DHx_w.jpeg" /></figure><p>Then, you’ll be prompted to enter your email — whatever you enter here, when you click the next step, your email is passed into the URL and subsequently leaked to the 3rd party advertising and analytics companies:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8kjdOHEU_t41RkRq5VIlig.jpeg" /></figure><p>Here’s a screen shot of the next step, with the user email being passed into the URL — the icon showing “45” is the Ghostery.com count of advertising &amp; analytics companies receiving data on the webpage — it’s not a complete list but this shows dozens of companies are receiving user emails from the current JetBlue.com data leak.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ydrCciVp4-YcOrpi5F1kxg.jpeg" /></figure><p>Here’s a screen shot from a previous test showing one of the advertising pixels firing and how it receives the user data through the request headers (notice only 39 pixels were tracked on this page last month, April’s test showed 45):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F0cDqoClUCJfjojJMOd1EA.jpeg" /></figure><p>The companies receiving data from Jetblue includes basically all the major advertising companies — Google, Facebook and all the niche but major advertising players. The Jetblue user email leak easily syncs to the most 3rd party companies out of any other leak in this research.</p><p>The Way Back Machine also has many copies of JetBlue’s website — at some point in 2019 they rolled out a new version of their website, and since at least July 9, 2019 they’ve been using the current version of their new account signup flow, you can see and literally test the July 2019 archived version here: <a href="http://web.archive.org/web/20190709195758/https://trueblue.jetblue.com/enroll/join-us">http://web.archive.org/web/20190709195758/https://trueblue.jetblue.com/enroll/join-us</a></p><p>Here’s a screen shot of the July 2019 version of the JetBlue.com user account creation 2-step form that leaks user emails on the 2nd step:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Hkiqpk8N-WIaExFYl9UBTA.jpeg" /></figure><p>Jetblue.com has been leaking user emails for about nine months for people creating new accounts…. it’s unclear when JetBlue will update this but they have rejected the research even though being informed on multiple occasions.</p><h3>The Bezos-Schmidt-Funded KongHQ.com (Formerly Known as Mashape) Using Common 2-Step Form That Leaks on the 2nd Step</h3><p>The company formerly known as Mashape, now known as KongHQ, was founded in 2007 and received $1.5 million in seed funding in 2011 from a <a href="https://www.eweek.com/development/google-s-schmidt-amazon-s-bezos-help-fund-mashape-api-marketplace">round of investors that included Jeff Bezos and Eric Schmidt </a>through Innovation Endeavors.</p><p>KongHQ has a 2-step signup form similar to the JetBlue leak, but the KongHQ form starts on their homepage. When a user puts their email in the form on the homepage and hits enter, their email is immediately pushed into the URL bar and then transmitted to 3rd party advertising and analytics partners.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z3ClD68LiPRIDaPw6FC4wg.jpeg" /></figure><p>KongHQ was informed of this breach back in February 2020 but still haven’t made any changes to their website and 2-step form — their response was similar to JetBlue in totally ignoring the issue.</p><p>In the original tests KongHQ transmitted data to:</p><ul><li>Google</li><li>Linkedin Twitter</li><li>Facebook</li><li>Drawbridge</li><li>Mixpanel</li><li>CrazyEgg</li><li>New Relic</li><li>Pardot</li><li>Wistia</li></ul><p>You can breach your own email address right now by filling out the form on the homepage of KONGHQ.com, but user beware!</p><p>After clicking “Request Demo” on the homepage, you are transmitted to the 2nd step of the form, with your email address added into the URL bar to auto-fill the form…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0YwiXZ4HLePuhpFAn_p0Tw.jpeg" /></figure><p>Unfortunately, anywhere you can find a 2-step signup form where the 2nd step has some form of autofill, many of those systems are being built with insecure technology and sometimes the user emails are purposefully leaked to optimize retargeting advertising campaigns or improve analytics attribution data.</p><h3>Democratic Data Broker NGPVAn.com / EveryAction.com (&amp; Their Clients) have been Pushing User Emails into Google Analytics &amp; Other Systems for Years</h3><p>NGPVan.com/EveryAction.com are owned by the same company and provide a wide range of CRM/marketing services for political and nonprofit clients. These platforms have an enormous range of features — and similar to the Mailchimp-Mandrill email breach described in this research, NGPVan created a legacy URL field for “emailAddress” that is appended into URLs, mostly on unsubscribe pages, and this can lead to NGPVan/Everyaction clients leaking user emails to 3rd party advertising and analytics companies.</p><p>A typical NGPVan unsubscribe URL that has the user email in it, looks like this (The email is appended at the back of the URL):</p><p>email.everyaction(.)com/unsubscribeUnique/3d7e893c-921a-ea11-828b-2818784d6d68/ad179bfc-9c1a-ea11-828b-2818784d6d68?nvep=ew0KICAiVGVuYW50VXJpIjogIm5ncHZhbjovL3Zhbi9FQS9FQTAwMS8xLzU1NDA3IiwNCiAgIkRpc3RyaWJ1dGlvblVuaXF1ZUlkIjogImFkMTc5YmZjLTljMWEtZWExMS04MjhiLTI4MTg3ODRkNmQ2OCIsDQogICJFbWFpbEFkZHJlc3MiOiAiemFjaEB2aWN0b3J5bWVkaXVtLmNvbSINCn0%3D&amp;hmac=73-vUhguqpitg-5DybUg7PmTNqxOTTLllnLe8CYE0y0=&amp;id=107423875&amp;<strong>emailAddress=cats%40victorymedium.com</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RujxgvKaodDT0PJk0iHelw.jpeg" /></figure><p>Unfortunately, not only are advertising and analytics companies ingesting the user emails on random unsubscribe pages all across the NGPVan client base, but those same URLs with user emails in plain text are also cached in <a href="https://www.google.com/search?q=site:everyaction.com+inurl:emailAddress&amp;filter=0">Google search results</a>, <a href="https://urlscan.io/search/#ngpvan.com*%20emailAddress*">URLscan results</a>, and in other repositories of cached user pages across the internet.</p><p>The primary company ingesting the user emails from NGPVan clients appears to be Google via their Google Analytics product, but a Microsoft endpoint also receives data. Here’s what Ghostery picks up on one of the unsubscribe pages with a user email in plain text in it:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uw4UBdqg-6WY1RkunNjl4g.jpeg" /></figure><p>And then here’s an additional screen shot of the actual data transfer, showing that Microsoft is also receiving the user emails through their visualstudio.com endpoints.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xQaVjnok8hiZ9XzbHIzhNA.jpeg" /></figure><p>NGPVan has been appending the user email address to unsubscribe links across their own emails, and client emails for several years — the start date isn’t exactly clear but emails from 2018 have this same plain text email.</p><blockquote>Google has also been aware of the NGPVan user email leaks since January, and Google clarified their Google Analytics policy around this type of ingestion, which apparently requires a <strong>“mandatory remediation process with the customer where they must stop sending PII to Analytics and ensure all historical PII data must be removed.”</strong> This statement was sent by Google on January 13, 2020, and several organizations have been flagged for Google who are sending user emails into Google Analytics, yet it appears no actions have been taken by Google on any of these issues.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pxAHUPipseB2717dbntGEA.jpeg" /></figure><p>NGPVan is currently used on the <a href="https://covid19responsefund.org/">https://covid19responsefund.org/</a> website sponsored by the World Health Organization (WHO), the United Nations Foundation, the Swiss Philanthropy Foundation, and <a href="https://covid19responsefund.org/supporters">with supporters including Google, Facebook, Microsoft, and others</a>. It’s unclear if people who donate money through the NGPVan donation form on the website are subsequently sent emails with their email address leaking via the unsubscribe links, but the form does provide options to join the email lists of several sponsoring organizations…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jwsgEPYmFQrgiaxt9eE1zw.jpeg" /></figure><p>NGPVan is a for-profit company and just because their clients are largely political campaigns and nonprofits, it doesn’t give them the right to leak user emails to advertising and analytics companies — hopefully this issue is resolved so that as the 2020 campaign heats up and users take advantage of unsubscribe forms more often, those user emails aren’t also leaking en-masse to 3rd party companies.</p><h3>Growing Child, Popular Magazine for Parents, Leaking Emails on Unsubscribe Page to Google Analytics, Google’s DoubleClick, only Google Pixels Receiving Data</h3><p>GrowingChild.com is a magazine founded in 1971 that describes itself as “serving millions of families in the United States and around the world…”</p><p>Unfortunately for the families who have subscribed to GrowingChild.com newsletters and then decided to unsubscribe, their unsubscribe pages print the user email in plain text into the URL and then share the user email in plain text to Google and several Google products including Google Analytics, Google Doubleclick and several other Google advertising endpoints.</p><p>The GrowingChild unsubscribe URls are built like this: <a href="https://growingchild.com/index.php/unsubscribe/unsubscribe.html?email=growingchild@victorymedium.com">https://growingchild.com/index.php/unsubscribe/unsubscribe.html?<strong>email=growingchild@victorymedium.com</strong></a><strong> </strong>and this leaks as a referrer to the Google pixels here:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FKQbnQYubpHnTSeVaa_sMQ.jpeg" /></figure><p>It’s unfortunately too common for unsubscribe pages to be built this way from legacy organizations, but organizations like Google seem to almost capitalize on it sometimes, like in the requests above that trigger across numerous Google advertising endpoints.</p><h3>MailChimp’s Mandrill Legacy Email Redirect via their API Can Leak Mandril-Client-User Emails to Advertising and Analytics Companies</h3><p>Mailchimp’s developer product Mandrill.com was <a href="https://www.cmswire.com/digital-marketing/mailchimps-mandrill-move-enrages-email-users/">founded in 2012 and claimed 80,000 users by 2015</a> — lots of developers still use their products, but one of their legacy APIs still has some clients using it. This legacy Mandrill API has a feature that can be used and then it can potentially expose user email addresses on unsubscribe pages to 3rd party advertising and analytics companies.</p><p>This Mandrill API doesn’t automatically leak user data but there is the option for Mandrill clients to redirect an unsubscribe URL sent via their API to include the user’s email address in the URL bar.</p><p>Mailchimp was informed of this issue relatively recently, they acknowledged the report and mentioned it was already escalated, but haven’t appeared to make many changes yet besides deleting old support articles which recommended the process that could potentially leak a user email to a 3rd party advertising or analytics company.</p><p>Here’s an example MailChimp Mandrill API redirect URL; this will redirect into a business email address for a newsletter:</p><p><a href="https://launch.us2.list-manage.com/track/click?u=baefb9fcb23d26e0308254e5c&amp;id=87ad1425d5&amp;e=af88ddc5b8">https://launch.us2.list-manage.com/track/click?u=baefb9fcb23d26e0308254e5c&amp;id=87ad1425d5&amp;e=af88ddc5b8</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zHQ5CjwwvsTRAt6HQTMRNA.jpeg" /><figcaption>If you visit “list-manage.com” you’ll be redirected to a Mailchimp error/details page</figcaption></figure><p>This “list-manage(.)com” domain is owned by MailChimp and numerous legacy Mandrill clients can be found who have various endpoints from this domain embedded and cached in URLscan.io like in this screen shot..</p><p>Google has also cached ~47,000 of the Mandrill unsubscribe pages via this search @ <a href="https://www.google.com/search?q=site%3Alist-manage.com%20unsubscribe">https://www.google.com/search?q=site%3Alist-manage.com%20unsubscribe</a> — not all of these results have user emails appended to them, which makes it clear that this is not a feature deployed by all Mandrill clients.</p><p>The process to add user emails into the Mandrill redirect URLs was covered in a legacy support article from Mailchimp. This support article was sent to Mailchimp with this research that they haven’t substantially responded to, yet they had time to delete the support article and try to hide what they were recommending to their clients — this page was deleted: <a href="https://mandrill.zendesk.com/hc/en-us/articles/205583017-Can-I-add-an-automatic-unsubscribe-link-to-Mandrill-emails-">https://mandrill(.)zendesk(.)com/hc/en-us/articles/205583017-Can-I-add-an-automatic-unsubscribe-link-to-Mandrill-emails-</a></p><p><a href="https://mandrill.zendesk.com/hc/en-us/articles/205583017-Can-I-add-an-automatic-unsubscribe-link-to-Mandrill-emails-">Can I add an automatic unsubscribe link to Mandrill emails?</a></p><p>Even though Mailchimp deleted this support article sometime in the last week, it’s still available <a href="https://webcache.googleusercontent.com/search?q=cache:lProQZp0dSoJ:https://mandrill.zendesk.com/hc/en-us/articles/205583017-Can-I-add-an-automatic-unsubscribe-link-to-Mandrill-emails-+&amp;cd=3&amp;hl=en&amp;ct=clnk&amp;gl=us">in the Google Cache</a>, you can see how Mailchimp showed how to put the user email into a specific query string:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Rn9cQ1SYM6MrU21o87iDJA.jpeg" /></figure><p>MailChimp also scrubbed their larger support article in Mandrill about unsubscribe pages which used to be @ <a href="https://mandrill.zendesk.com/hc/en-us/articles/205582947-About-Unsubscribes">https://mandrill(.)zendesk(.)com/hc/en-us/articles/205582947-About-Unsubscribes</a> — the page is about appending user emails into the URL bar —<a href="https://webcache.googleusercontent.com/search?q=cache:3aTbGl6bjNUJ:https://mandrill.zendesk.com/hc/en-us/articles/205582947-About-Unsubscribes+&amp;cd=1&amp;hl=en&amp;ct=clnk&amp;gl=us"> Google cached that page too here</a>.</p><p>Here’s a screen shot of that page before Mailchimp deleted it this past week:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J1t-skeRlDjHk3w00v6DXQ.jpeg" /></figure><p>Again, MailChimp received the report, they’ve obviously been scrubbing their content since receiving the research, but haven’t sent any other details about their plans to notify users or rearchitect the Mandrill service.</p><h3>Washington Post Leaks Some User Emails in Base64 to Service Providers, Appears Not to Send Data to Any External Advertising Companies</h3><p>The Washington Post was recently alerted to a base64 user email data leak to a limited number of analytics companies, primarily Chartbeat.com (and maybe a few others) — it appears no advertising companies received the base64 user email strings that several of their newsletters append to their unsubscribe links.</p><p>The Washington Post escalated the initial report quite fast and noted they were addressing their issues — Wapo’s base64 user email sharing could be resolved by the time of publication or likely sometime soon after.</p><p>Here’s one of the unsubscribe links that has had the user base64 email strings:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dOm7T-eVbW-PqmFjIrfoiw.jpeg" /></figure><p>Not all of the Washington Post newsletters are built the same way — the leak occurs in the unsubscribe links for the “This Week in Ideas” newsletter and another one of their weekly newsletters — their core system for newspapers subscribers that sends daily emails does not seem to be built this same way and doesn’t seem to leak user emails.</p><p>The user emails are encoded in base64 plain text format and appended into a “bem” URL parameter — you can see one of these unsubscribe link via this link @ <a href="https://s2.washingtonpost.com/1d1e5c5/5e9c2f85fe1ff6038c0b8365/ZWR3YXJkcy56YWNoLnNjb3R0QGdtYWlsLmNvbQ%3D%3D/62/66/56455e06f2aadef382a0683ad4c5b64e">https://s2.washingtonpost.com/wp-unsubscribe/newsletters?bem=<strong>ZWR3YXXXXXXXXXNoLnNjb3R0QGdtYWlsLmNvbQ</strong>%3D%3D&amp;nlsendid=5e6e0c0bfe1ff6038cda4f2e</a></p><p>The base64 string is the “bem” param above — mine is slightly obscured above but these can be found across old emails and in some locations on the internet.</p><h3>Facebook Manipulates URL Query Parameters (for Filtering), But Still has System That Can be Broken to Leak Emails to 3rd Party Advertising and Analytics Companies</h3><p>Nearly all modern websites use Javascript for advertising and analytics tracking, but it’s still very rare for organizations to “sandbox” their partner javascript pixels in a way that ensures that URL parameters don’t get transmitted accidentally to ad tech and analytics partners.</p><p>Facebook is one of the few organizations that regularly does “URL referrer filtering” for their javascript partners, and across most (or all?) of their websites. Facebook does this to filter certain URL parameters like the <a href="https://medium.com/@thezedwards/facebook-s-ongoing-vip-user-data-exfiltration-vulnerability-via-adobes-marketo-software-why-d8435a259b0">“mkt_tok” parameter reported to them back in 2019 that could leak user emails through Adobe’s one-token-user-authentication architecture</a>.</p><p>Facebook has never open sourced this filtering product of theirs, likely because it helps to cut down on spam and extra work they need to do internally, but their filters are largely built on their own domain, and they have certain Facebook business marketing pages with 3rd party advertising and analytics pixels, and even though Facebook’s filtering product is deployed there, it still typically transmits user emails in some field into 3rd party systems — possibly in fields that can be easily purged, but ingested nonetheless.</p><p>Attached below is a screen shot example showing a type of filter where the unknown/unexpected URL parameters that I put into the Facebook URL were then passed into a unique Adobe Marketo field named “_mchQp” — it’s possible that this is a field Facebook parses to ingest unknown inbound data that may/may not get deleted based on some other criteria.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5ehWPSA_NfLLJQtD2ZZ-XQ.jpeg" /></figure><p>Organizations that have Javascript advertising and analytics partners need to be aware of their own user data breaches but also plan for ways that attackers could inject bad data into 3rd party systems to corrupt retargeting campaigns or break analytics systems. And at some point, more organizations will need to look to architecture from orgs like Facebook who filter URL parameters and build internal sandboxes to protect user data from flowing across their global data supply legal exposure chain.</p><h3>What’s Next? What Questions are Important?</h3><p>The organizations included in this research can request any changes or comment additions via URLdatabreach@victorymedium.com</p><p>Individuals who use any of these services and who believe had their emails leaked, should be given an easy processes to request the deletion of their user emails that were sent to 3rd party advertising and analytics companies. The organizations involved in this research should provide that process in whatever format they can provide.</p><p>Each organization included in this research should be changing their systems to stop leaking user emails in plain text or base64 plain text formats, they should notify users who could have been impacted by the leaks, and also issue deletion requests for all their users to 3rd party advertising and analytics organizations they work with.</p><p>All organizations should be extremely careful about 2-step forms, email tracking that appends encoded or plain text emails into URLs, and any process that “syncs a user email” to a 3rd party company. This process is almost assuredly not described properly in Terms of Service and Privacy Policies for organizations, and it’s obviously not a process that most users expect to occur.</p><p>Unfortunately, as auditors saw with the Cambridge Analytica scandal and Facebook’s inability to confirm that the data was completely deleted, the organizations involved in this research face a similar dilemma tracking down and deleting user emails that were sent to their 3rd party advertising and analytics partners — how can you actually ensure and know this data was deleted? How can users who were involved in these flows ensure their emails are deleted? Who is in charge of requesting that? Each user to every ad tech/analytics company? Each user to the original offending organization? Should organizations proactively request the deletion for all their users? Will the previously leaked user email data just stay with these 3rd party advertising and analytics companies with only a small minority of users requesting deletion?</p><p>Finally, many advertising companies have features they’ve built to sync user emails into retargeting lists and other audience advertising targeting strategies, without properly notifying users? How many of those organizations have user emails that were given without the user fully understanding what was occurring or having an ability to delete or modify that information after it was sent?</p><p>Hopefully, organizations will start to take a more proactive approach to trying to stop this type of data supply data breach, and a more responsible plan of action after being notified of significant problems.</p><p>Additional questions or concerns? Ping me on twitter @<a href="https://twitter.com/thezedwards">thezedwards</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=39a09d2303d2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What your lawyers say you’re doing VS What your growth team and developers are actually doing ….]]></title>
            <link>https://medium.com/@thezedwards/what-your-lawyers-say-youre-doing-vs-what-your-growth-team-and-developers-are-actually-doing-a0472a5a67e9?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/a0472a5a67e9</guid>
            <category><![CDATA[gdpr]]></category>
            <category><![CDATA[user-privacy]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[ccpa]]></category>
            <category><![CDATA[user-data]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Wed, 01 Jan 2020 19:09:14 GMT</pubDate>
            <atom:updated>2020-01-01T19:09:14.970Z</atom:updated>
            <content:encoded><![CDATA[<h3>What your lawyers say you’re doing <strong><em>VS</em></strong> What your growth team and developers are actually doing …. the real impacts of CCPA and GDPR</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ph2mzt-h5AUi6yb43DWSnw.png" /></figure><p>CCPA and GDPR force companies to “put pen to paper” about their global user data policies and partner data sharing, but as with any system of accountability in a marketplace, the mere existence of regulatory frameworks that force transparency, don’t ensure that the data being shared is accurate or universal for all users.</p><p>And with both CCPA and GDPR, there will always be a common mantra of exposure… <strong>“your data sharing partners put your business out of compliance…”</strong></p><p>Over the last 12–24 months, numerous companies have been founded that probably have on the first slide of their pitch decks something along the lines of, “We’re going to be LegalZoom™ for User Data Privacy Requests and Complaints!” These are typically data supply products but not data supply solutions –they are the equivalent of a Spiderman Band-Aid for a double amputation.</p><p>I’ve personally vetted most of the new data supply companies that have been popping up to support American user data privacy complaints, and I also did the same before/during/after GDPR with mostly-European companies founded to support the IAB’s Consent Management Provider frameworks.</p><p>When I vetted the CMP’s, I found that the very best CMP was a little-known (at the time) tool called “Faktor.io” and they were special for a number of technical reasons that I don’t need to share, and their company was eventually acquired by the massive LiveRamp for an undisclosed sum. Similar situations exist with how companies are approaching CCPA and the whole concept of user data compliance and revocation audits — some companies are doing it right, and others are just “going through the motions” and trying to copy more sophisticated peers.</p><p>GDPR forced companies across the world to think about “cookie consent boxes” not as a UI/UX legal catch-all, but as a technical barrier, controlled by the user, which could actually, factually, segment a user’s data flow based on their preferences, and also support the revocation of that data along the supply chain, as-warranted by the user at some point in the future or based on pre-agreed-upon rules.</p><p>Good CMPs are not really about “managing cookies” but much more about federating strings and organizing the fine-grained event data so that the CMP was more like a tag management solution with 3rd party integrations and revocation triggers — a full-cross-platform-data-event-suite. Bad CMP’s are just rudimentary page-load JavaScript blockers with bolted on post-consent JS piggybacking. The difference is vast, but woefully misunderstood by numerous companies in the market today.</p><p>Below you’ll find a flowchart I made a few years ago that demonstrated where holes could exist in a user data consent flow across devices and user experiences. If your organization conducts a real and similar audit and flowchart across your digital platforms , you will immediately find huge user data holes across various features, apps, integrations, widgets, and partners in your ecosystem — guaranteed.</p><p>There are technical ways to do parts of these audits, but data supply auditing has no silver bullet — you must carefully audit every piece of technology, every user step, every engagement, and map what is actually, factually happening. Otherwise, if you rely on JavaScript CMPs and other technical “user data blockers” without going that extra step, you’ll never see what data you aren’t collecting and what data is flowing outside your control.</p><p>Where is the upstream user data noncompliance for your organization? Where is user data shared, and where do your partners have the opportunity to share it even further outside your control?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vvLU5bbo7pzlT0oxFTZZCw.png" /><figcaption>Example data supply audit of an architecture using Segment.com and various partners // System maps like this should be expanded until 100% of all technology is covered, with systems to identify new technology or partners across a company’s entire technology stack (not just the website).</figcaption></figure><p>GDPR, for all its perceived faults, created very strong definitions for “data controllers” and “data processors” — with the primary difference being that if you are a “controller” you are responsible for the data that you share to all of your partner processors — and if one of those processors then shares to additional/new companies, that processor is upgraded to a controller, and the data supply chain now has multiple joint-controllers of data, with their own legal exposure to complaints.</p><p>CCPA’s framework is obviously different than GDPR, but there are numerous similarities with how it forces companies to “identify partners” and have the ability to control the flow of the user data to those partners.</p><p>CCPA also has many months ahead before it will be clear how enforcement plays out, and with the impending second data privacy referendum being put on the California ballots in 2020, and potential efforts to pass a wide-ranging user data privacy and security bill in Congress, we’re only dipping our toes into what will eventually become a working user data and user PII framework here in the United States.</p><p>For decades there have been fragmented efforts in the U.S. on these, from healthcare privacy to banking privacy, to video game rental privacy, to mobile phone beacon privacy, and other niche-privacy efforts that largely just created a patchwork of confusion that is rarely enforced by the FTC and other relevant agencies.</p><p>The private right to action in CCPA and other American user data privacy frameworks that are coming online in 2020 (and potentially strengthened by Congress) are going to further differentiate the “American GDPR” vs the actual European GDPR — and we’re many years from companies across the internet ecosystem fully feeling the effects of all these changes.</p><p>In the coming months and years, some companies will lurch out of existence quickly from enforcement actions or scandals, whereas others will slowly lose revenue as data flow stops to their previously-non-compliant user data sharing practices.</p><p>And as someone who has been building and optimizing data flow systems for over a decade, and auditing complex data supply projects for several years, I know for a fact that I don’t always have the answers, because every technical system has new ways to share data and new protocols that partners can exploit, and every company has partners and features/apps/integrations/widgets that only their marketing team or development team knows about — and due to the inherent nature of running a business on the internet, in any given week, it will be your developers and marketers who put your entire business out of compliance. And your exposure will be further compounded by your partners sharing data with their partners, further outside your control.</p><p>Moving forward, when it comes to data supply auditing and user data privacy compliance frameworks, there will always be a gap between what your lawyers say you’re doing, versus what your growth team and developers are actually doing.</p><p>2020 is going to be the <strong>first</strong> year that U.S. companies finally know what happens when they can’t account for their user data or their partner practices… but there are years of claims, settlements and FTC consent decrees ahead for those unwilling to acknowledge this shift.</p><p>//</p><p><a href="https://twitter.com/thezedwards">Follow me on Twitter @ thezedwards</a> for updates in the weeks to come on CCPA, GDPR and important data supply research about the global ecosystem.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a0472a5a67e9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ClearCoin Google Chrome Extension Hijacking Online Ads — Captured Data Appears to Show Amazon…]]></title>
            <link>https://medium.com/@thezedwards/clearcoin-google-chrome-extension-hijacking-online-ads-captured-data-appears-to-show-amazon-79d046f0502c?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/79d046f0502c</guid>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[malware]]></category>
            <category><![CDATA[ad-fraud]]></category>
            <category><![CDATA[advertising]]></category>
            <category><![CDATA[chrome-extension]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Sun, 07 Apr 2019 23:51:42 GMT</pubDate>
            <atom:updated>2019-04-07T23:51:42.887Z</atom:updated>
            <content:encoded><![CDATA[<h3>ClearCoin Google Chrome Extension Hijacking Online Ads — Captured Data Appears to Show Amazon Buying the Hijacked Impressions</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1I5P2n7feqBPuYk0jNzVyQ.jpeg" /><figcaption>Watch the video of this hijacking at the bottom of this post…</figcaption></figure><p>ClearCoin was one of the first Escrow Tokens on the market — and it was originally founded and supported by the right-hand-man of Satoshi Nakamoto, <a href="https://twitter.com/gavinandresen">Gavin Andresen</a> (a mega-genius in his own right who thinks outside the box with cryptography solutions).</p><p>In June 2011, ClearCoin was put on ice by Gavin, who said he couldn’t give it enough time and wanted to focus on Bitcoin. <a href="https://web.archive.org/web/20140518040025/https://bitcointalk.org/index.php?topic=21659.0;all">The forum thread is archived here on the Way Back Machine</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eBlYwJCjJYSpp0iKfnD2yg.jpeg" /></figure><p>The ClearCoin and Bitcoin community largely seemed to understand, but also many were requesting the option to purchase the ClearCoin business from him — it was a novel Escrow DApp. He spoke many times about how he was willing to sell ClearCoin, but needed to know that it would be in safe hands…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3sf-UgrZpIlCmjwXnfnizw.jpeg" /></figure><p>Between those posts in 2011 until now, the ClearCoin business timeline gets a bit murky (although it seems to have been sold at some point)….so i’m going to skip to 2018–2019 and the data that is currently available…</p><p>Fast forward, and the ClearCoin escrow technology is not only still being used, but it was turned into a glorified “Demand Side Platform” for online advertising — basically the new owners took the concept of publishers demanding refunds (Escrow), and duct-taped some logic together to let publishers and advertisers theoretically create advertising agreements where one party could actually not pay based on a disagreement — holding the transaction back from being completed and forcing some sort of off-chain arbitration to complete the transaction.</p><p>This is how the logic for ClearCoin was originally/currently described over at Wikipedia (which likely has been updated in some ways):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*W-hbe6kxSqqW8hswceAz8A.jpeg" /></figure><p>Now, most folks in eth/bitcoin or the online advertising industry must be scratching their heads wondering…. <strong>so who owns ClearCoin now? Who is powering this Demand Side Platform? What does this have to do with a Google Chrome Extension?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ADggmYG37yuhhodukDWquw.jpeg" /></figure><h3><strong>Beeswax.com Powering the Demand Side Platform ClearCoin</strong></h3><p><a href="https://clearcoin.co/company/telegram-community-qa/">In August 2018, ClearCoin posted a FAQ</a> to help answer questions from the community — during this process they were asked, “Have any companies actually started using the platform yet and if so, have the team had user feedback?”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Stetf3fnDmhUG3PAtKCnMQ.jpeg" /></figure><p>ClearCoin responded, writing:</p><blockquote>“…<strong>We are working with </strong><a href="https://beeswax.com/"><strong>Beeswax</strong></a> for supply side access to ad inventory. The company provides hundreds of millions of queries per day on major supply sources. We’re extending our conversation with them as they may be able to provide demand for the exclusive supply side inventory that we secure. There have been a few other pilots that we cannot name yet due to non-disclosures. The intention is indeed to pursue, secure, and announce as many major partnerships as possible and that includes household names. Stay tuned to see what is announced in the future.”</blockquote><h4>Beeswax.com — Bidder as a Service &amp; ClearCoin Tech Vendor</h4><p>Beeswax.com is an online advertising demand-side platform (DSP) that was originally founded in 2014. Beeswax has described themselves as A “bidder-as-a-service” business — if you don’t understand — think of them as a middle-layer between a publisher and other established ad networks — that’s probably the simplest description.</p><p>The Wall Street Journal covered their work early in 2019 and somehow left off the fact that they were the tech infrastructure for ClearCoin — the article reads more like a glorified press release but <a href="https://www.wsj.com/articles/ad-tech-in-housing-firm-beeswax-raises-15-million-11548154801">you can read it here</a>. Their round and previous funding was described as:</p><blockquote>Beeswax.io Corp, an ad-tech firm that helps marketers build their own in-house digital ad-buying capabilities, <strong>has raised $15 million in a new funding round</strong>, the company said, another sign of the growing interest in marketers seizing more control of their advertising function.…Beeswax’s latest round, at a $77 million pre-money valuation, was led by venture-capital firms RRE Ventures, Foundry Group and Amasia. <strong>Beeswax last raised $11 million, at a $24 million pre-money valuation, in 2016</strong>. Founded in 2014, the startup has now raised $28 million in funding and has 65 employees.</blockquote><p>The timeline here makes it pretty clear that Beeswax last raised $11 million in 2016, potentially before the big partnership and launch with the newly improved ClearCoin. And then a few months after the ClearCoin Google Chrome Extension launched, Beeswax raised another $15 million.</p><p>Now, in terms of ClearCoin’s direct funding, I haven’t found an exact source for how much they raised during their Initial Coin Offering (ICO) but in February 2018 it was at least $3 million, <a href="https://www.newsbtc.com/2018/02/05/ad-specialist-startup-running-ico-successfully/">according to an editorial about CC from NewsBTC.com</a>.</p><p>Since that time and the initial raise, the price of ClearCoin has gone into the toilet like most ICO’s, losing 84% of its value in the last 12 months. But the original ICO money was obviously spent on R&amp;D….</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2Nnub14r8JX34UUgsdNsxQ.jpeg" /></figure><p>Now you may be asking yourself, is Beeswax.com still the engine driving ClearCoin? And if-so, what proof is there? DNS records are public and supporting technical infrastructure creates clear ties between businesses.</p><p>This is the login page for ClearCoin, hosted by Beeswax: <a href="https://securitytrails.com/domain/clearcoin.beeswax.com/dns">clearcoin.beeswax.com</a></p><p>Many important questions remain about Beeswax and their relationship with ClearCoin…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XE8XGQMRqWDXaN3eNgCGiA.jpeg" /></figure><h4>ClearCoin Operating with New Google Chrome Extension Hijacking Online Ads Since Fall 2018</h4><p>ClearCoin launched a Chrome Extension in Fall 2018 that replaces ads on other websites with the ads from ClearCoin…. AKA…. whatever company provides the ads for the ClearCoin network (all signs point to Beeswax tech), is using a Chrome Extension to replace publisher ads with their own ads. This steals money from publishers and other ad networks and is explicitly banned in pretty much all advertising Terms of Service (TOS).</p><p><a href="https://chrome.google.com/webstore/detail/clearcoin-the-ad-blocker/benncgglohdbeapbakcebdobkdbkdcba">Here is the current public Chrome Extension</a> (be careful about installing), and screen shots:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HC_4-qy9yOn12HfWm2x1Ng.jpeg" /></figure><p>“The ClearCoin Extension <strong>replaces the normal ads you see while browsing</strong> <strong>the web with ads from our network</strong> that reward you, the user, with XCLR tokens.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0-eM5wxcukIjh5GbI4sfYw.jpeg" /></figure><p>Here’s the ClearCoin Google Chrome extension permissions that’s it’s requesting :</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*weE5TCzm8mEiu5QtHu7TVg.jpeg" /></figure><p>Now, Google has been fighting back against <a href="https://www.pcworld.com/article/2904852/google-cracks-down-on-adinjecting-chrome-extensions.html">Chrome Extensions that inject advertising since 2014-2015</a> — and they’ve removed hundreds of extensions that abused permissions to hijack publisher banner revenue.</p><p><a href="https://portswigger.net/daily-swig/google-chromium-changes-may-break-ad-blockers">In January 2019, Google even announced more changes to Chrome Extensions</a> that could break “Ad Blockers” — the changes were heavily focused on any extensions that utilized the “webRequest API” to hijack advertising requests — either for injecting their own ads or for straight up blocking. Google’s policy here continues to evolve with lots of public feedback, and it’s important for technologists to parse these details and provide feedback.</p><h3>Watch the Video of the ClearCoin Chrome Extension Hijacking Banner Ads and Trying to Inject Amazon Ads</h3><p>Here’s a video showing how ClearCoin hides ads on a publisher website and tries to inject their own ads. In the video you’ll see that the only ads injected into the test publisher site were associated with Amazon products. It’s unclear why this publisher was limited to Amazon, but I could hypothesize that it’s largely because Amazon is aggressively buying on almost every network out there and may be too swamped to audit fake impressions effectively.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FXrllMPfavOk%3Ffeature%3Doembed&amp;url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DXrllMPfavOk&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FXrllMPfavOk%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/b66da2ee440699170455b1ecc10995a6/href">https://medium.com/media/b66da2ee440699170455b1ecc10995a6/href</a></iframe><p>Here’s a screen shot of the bid requests — the creative is served from: <a href="https://s3.amazonaws.com/clearcoin-creatives/*"><strong>https://s3.amazonaws.com/clearcoin-creatives/*</strong></a></p><p>You’ll also see in the screen shot below that immediately after the ads were injected/served from the AmazonAWS instance, Google fired off some requests capturing the details of the ad injection…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5zVYMgmYcYYBPd9B4FRXhw.png" /></figure><p>Here are three of the banner ads that ClearCoin tried to inject — notice they are all Amazon ads — so far I haven’t found any non-Amazon ads served through this, but testing was limited to one session:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*2wCG3As5U5nEjyeCYIGLhA.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*bFa3GMAjc5TB6Nnbg2voVQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*DHYk0TncfR1C4R14VQoOyA.jpeg" /></figure><h3>Google Caught the Injected Ads from ClearCoin &amp; Sent Data to a Google Reporting Tool</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5zVYMgmYcYYBPd9B4FRXhw.png" /></figure><p>After those creatives were served, a unique thing happened — you can see it in the video and screen shots to the left — but basically, as soon as one of those injected ads appeared, they would flash into view, disappear, and then Google would send a /SafeBrowsing/ClientReport request … with the HTML of the Amazon ads being sent to Google…</p><p>Aka… Google has the data internally to see that a malicious Chrome Extension is injecting ads, and they have copies of the creative served via the extension to real users ….</p><h3><strong>So what’s next? Why post this? Why does this matter? Are ad networks hopelessly broken?</strong></h3><p>In short, the online advertising industry is broken. <a href="https://twitter.com/thezedwards/status/1112563838178607105">I’ve written extensively about how piggybacking javascript for ad/audience networks is the subprime infrastructure of the internet </a>— but there are even darker corners of the online advertising industry that are being supported by so-called respectable advertising organizations.</p><p>In the coming months, it’s likely Google Chrome will not only continue to remove malicious extensions that hijack ads, but implement additional restrictions for 3rd party javascript and cookies.</p><p>If advertising companies think that their path to surviving the upcoming cookie-apocalypse runs through grey or black hat advertising strategies like revenue hijacking, they are very mistaken… change is coming, and it will make users more safe, protect advertisers from forms of ad hijack fraud, and ensure that publishers aren’t having their content un-monetized by malicious extensions.</p><p>If you’re buying ads from a network, do you trust them? If you’re buying from a publisher, do you know their javascript deployment partnership strategies? Do you only buy ads on publishers with valid Content Security Policies and iframe-injection protections?</p><p>… it’s time to start to develop opinions on these things…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/450/1*5qgbu1kRWQUby4wRRBYO2Q.gif" /></figure><p><strong>Questions? Edits? Send me an email @ spaghettiJS@victorymedium.com or on twitter </strong><a href="https://twitter.com/thezedwards"><strong>@thezedwards</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=79d046f0502c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Facebook‘s Ongoing VIP-User Data Exfiltration Vulnerability via Adobe’s Marketo Software & Why…]]></title>
            <link>https://medium.com/@thezedwards/facebook-s-ongoing-vip-user-data-exfiltration-vulnerability-via-adobes-marketo-software-why-d8435a259b0?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/d8435a259b0</guid>
            <category><![CDATA[facebook]]></category>
            <category><![CDATA[gdpr]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[javascript]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Wed, 02 Jan 2019 16:36:56 GMT</pubDate>
            <atom:updated>2019-02-06T21:51:45.699Z</atom:updated>
            <content:encoded><![CDATA[<h3>UPDATE: Facebook Fixes VIP-User Data Exfiltration Vulnerability via Adobe’s Marketo Software &amp; Why One-Token-User-Authentication Protocols are Unsafe</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/318/1*0ebaoPhQ3AREQ_ZCGnJ1_g.gif" /><figcaption>Facebook removed auto-fill functionality on forms that used Adobe’s Marketo “mkt_tok” value and took other steps to protect Facebook users from one-token-user-authentication user data exfiltration risks.</figcaption></figure><p>Update: On January 29, 2019, I heard back from Facebook (27-days after I initially published this research) and they informed me they had taken action with Adobe to <strong>turn off auto-fill functionality on their forms based on the mkt_tok URL parameter.</strong></p><p><strong>Facebook’s about-face on this type of URL authentication is a huge win for privacy-advocates concerned with One-Token-User-Authentication URL parameters and their impact on user privacy.</strong></p><p>Facebook also made sure their marketing subdomain @ content.fb.com was properly served over HTTPS (this is still an ongoing migration), they are evaluating adding text to not forward marketing emails, are working to potentially whitelist certain domains and other efforts to limit the “mkt_tok” URL token leaks, and Facebook even announced they were limiting the use of external resources (JS, CSS) on pages that provide auto-fill functionality and receive mkt_tok values.</p><p>Facebook also reversed their original decision to issue me an average-sized bounty for the research and sent over some other helpful tips for future submissions. All in all, Facebook seems to have taken nearly all of the suggestions in this original article still included below (and they seem to be working on a few of the more advanced strategies to protect users who leak their own tokens but wouldn’t provide details when I inquired further).</p><h3>Here’s the email Facebook sent to me announcing their changes:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XD5B5oqDRSmy-X4i9yLrMQ.jpeg" /></figure><h3>Facebook Adopted Fixes, Adobe Still Largely Silent</h3><p>I’m still a little surprised that Facebook made the changes that I suggested in this research, particularly some of the changes that could impact their opt-in conversion rates, which is a surefire sign that the Facebook Growth team is having to take a backseat to the privacy and security team when subtle conflicts arise.</p><p>I’m not however surprised that Adobe and their subsidiary Marketo have been largely silent. I’ve seen no messages to their clients, no steps to help their other non-Facebook clients implement security precautions, and it’s likely that the steps Facebook took to protect itself from Adobe’s Marketo software vulnerabilities are not being taken by other Marketo clients. If I hear differently, I’ll update this post. I’ll also continue to monitor the use of One-Token-User-Authentication URL parameters like “mkt_tok” and will urge more folks start to look into the importance of using multiple cookies and values to authenticate a user prior to any user data sharing.</p><h3>…Here’s the original research on Facebook + Adobe that I published on January 2, 2019, that led to Facebook’s changes at the end of January 2019….</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3049pqSqP8CPv1jRE9fS-w.jpeg" /><figcaption>Just one of the many folks who shared their “mkt_tok” token inadvertently.</figcaption></figure><p>Over the last 2 months, Facebook has quietly changed several ways they deploy Adobe’s Marketo email software across core Facebook marketing programs and websites in order to reduce the impact of a one-token-user-authentication user data vulnerability. But Facebook also continues to send Facebook users emails on a daily basis containing one-token-user-authentication vulnerabilities, and they continue to host websites and signup forms that are vulnerable to targeted user data scraping.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dOyAnVUMFGX18uUeysMUwA.jpeg" /><figcaption><strong>All of the links in this December 2018 Facebook for Developers Newsletter exposes a “mkt_tok” one-token-user-authentication URL variable in the final resolved URL.</strong> In numerous Facebook newsletters, Facebook links their users to 3rd party websites that trigger javascript pixels for 3rd party companies who capture the full URL, including the private “mkt_tok” user token appended to the URLs. Facebook also sends users to domains they control like F8.com, which passes the user URL and private token to the advertising networks for Google, Twitter, and Linkedin. Users also inadvertently share their own tokens, sometimes on Facebook.</figcaption></figure><p>Facebook’s user data vulnerabilities have existed for over a year, and there are Adobe Marketo forms built in 2017 that are <a href="https://go.fb.com/analytics-email-sign-up.html">still</a> <a href="https://go.fb.com/subscribe-blueprint-webinars.html">hosted</a> <a href="https://go.fb.com/audiencenetwork-newsletter-sign-up.html">on</a> <a href="https://go.fb.com/discover-growth-subscribe.html?ref=discover_growth">Facebook’s</a> <a href="https://go.fb.com/demo-signup.html">domains</a>. These web pages and the ongoing Facebook email strategy make it possible for a Facebook user to share a URL that can be used to expose their email address and other private user information.</p><p>One important caveat to this user data vulnerability is that it <strong>*does not*</strong> <strong>impact normal Facebook user accounts</strong> that are used by the ~2 billion people (newsfeed, messenger, ads manager, etc.). This user data vulnerability merely impacts the “Shadow b2b Facebook User Accounts” built by Facebook and hosted by Adobe’s Marketo email software that Facebook uses to engage industry experts and encourage them to utilize Facebook tools, and then more recently the vulnerability was extended to Facebook Workplace users due to a deep integration between Facebook Workplace and the Adobe Marketo email software.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/990/1*8EzKse0bqqbn8DR4YpPEhw.jpeg" /></figure><p>One additional caveat, Facebook has been notified about these user data vulnerabilities via their Whitehat disclosure program in a thread with me containing over 30 messages, initially started on October 29, 2018.</p><p>Facebook acknowledged the insecurities in Marketo in plain English on December 3, 2018, by writing “at this time sharing those links is not recommended,” but have continued to send out the links to users via email on a regular basis, and have kept dozens of other web pages hosting Marketo forms active that facilitate the data scraping vulnerability. Unfortunately, both sides of this user data vulnerability equation are active and ongoing.</p><p>Facebook has since closed the whitehat thread with me, marked it as “Informative” but have kept huge portions of the vulnerability active and ongoing. Due to Facebook’s lack of remedy and ongoing actions, it seemed important to disclose this to users so they can protect themselves. I let Facebook know I was going to disclose this via a public post and sent them one more update this week.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lxnwBKgTKZ22h-RRrE2q1g.jpeg" /><figcaption>All of the links in this Facebook for Marketers “Facebook IQ” December 2018 Newsletter exposes a “mkt_tok” one-token-user-authentication URL variable in the final resolved URL. If any of these URLs are shared publicly after a user visits them, that user is publicizing a token that can expose their private information.</figcaption></figure><p>Finally, Adobe was notified and confirmed receipt of this user data vulnerability existing in Marketo on September 25, 2018 (more on this and screenshots of the conversations below), more than a month before they completed the acquisition of Marketo on October 31st, 2018 and have taken no visible steps since then to alert clients to ongoing user data vulnerabilities or fix these foundational problems in their Marketo software.</p><p>This blog post is going to largely focus on Facebook’s deployment of Adobe’s Marketo email software, and why due to Facebook’s internal marketing team’s needs to “move fast and break things” and their need to flexibly control external email newsletters, a feature that (for some reason) doesn’t currently exist within Facebook’s tech stack, their reliance on Marketo created vulnerabilities across different Facebook divisions, external programs, and VIP user bases containing high-spending ad buyers and Facebook platform developers.</p><p>Unfortunately, by disclosing this one-token-user-authentication user data vulnerability, it also puts other Adobe Marketo clients at risk of targeted data exfiltration — but Adobe has experience fixing this problem with their own Adobe Marketing Cloud software (I disclosed a similar problem to them last Spring, more below on that too), and Adobe needs to take the same steps with their new acquisition Marketo.</p><p>Unfortunately for Adobe, it may be more complicated to deploy two-token-user-authentication within a legacy data architecture like Marketo due to Marketo’s heavy-reliance on one-token-user-authentication. Marketo uses the same user token to both track email opens and clicks, and also uses that same exact token to control updating and accessing user preferences and contact information. <strong><em>“One user token to rule them all,” if you will.</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6IQgE40gAXDPaQjfvXfVuA.jpeg" /><figcaption>All of the links in this December 2018 Facebook for Business Newsletter exposes a “mkt_tok” one-token-user-authentication URL variable in the final resolved URL.</figcaption></figure><p>In short, this vulnerability exists due to a string of text, a variable called “mkt_tok” — this string of text is 23 alphanumeric characters — it’s appended to the end of website URLs by Marketo’s email software after a user clicks a link in an email sent to them by a Marketo client.</p><p>That “mkt_tok” token controls both read/write access to the user data, and it also tracks the email opens/clicks. The “mkt_tok” user token is shared from Marketo’s email software, and it seems that nearly all email newsletters sent through their platform have this user token appended to URLs. This means that anytime Client A shares Website B with User C, if that user clicks to visit the website, the user token will be appended to the URL and Website B will see this private one-token-user-authentication variable in all their server logs and analytics tools — and that user token is also shared across the data supply chain to dozens or hundreds of advertising and analytics companies on some popular websites.</p><p>On multiple occasions, Facebook sent newsletter emails to their users that linked the user to a 3rd party website, while also appending the users private “mkt_tok” to the URL after clicking the link in the email, and in milliseconds of a user clicking an email, their private user token would pass to the 3rd party website and any advertising and analytics tools that are fired on that domain that grab the final resolved URL.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j-RHwuimYmYj9HHeBBMGag.jpeg" /></figure><p>This one-token-user-authentication leak also occurs when a user clicks to visit a web page, has their user token appended to their URL, <strong>and then shares that blog post or page they are visiting publicly — like by writing a public post on Facebook</strong>. This is the most common way that Facebook’s VIP users are exposing themselves, and it starts with them sharing Facebook Marketing and Developer news that they were sent via email from Facebook, back onto Facebook. This full loop of “promoting Facebook content on Facebook” exposes a users “mkt_tok” token publicly in the URL they share.</p><p>And unfortunately, Facebook also still lets any user <a href="https://www.facebook.com/search/posts/?q=&quot;mkt_tok&quot;">search Facebook to look for public posts that contain the URL variable “mkt_tok”</a> (as does Google) — so these user tokens are widely available in public locations. Facebook was notified about these issues but made only minimal changes before closing the issue, so it seems important to alert my fellow marketers and developers to this vulnerability so they stop sharing links sent via these Facebook programs. Facebook Workplace users should also be warned, due to the inclusion of these tokens in new-user Facebook Workplace onboarding email campaigns sent by Facebook.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dLlxCFD1CB8WvjJPgTiPBA.jpeg" /><figcaption><strong>This email from Facebook is safe. </strong>Facebook sends Developer email updates for new blog posts, and these are sent securely using Facebook’s simple email tool. The links in these emails *do not* contain any tokens or user data vulnerabilities.</figcaption></figure><p>Finally, this blog post will briefly touch upon the dangers of one-token-user-authentication, why user tokens are rightfully-protected in both Europe’s user privacy framework known as GDPR but also the more-recent California Consumer Privacy Act (CCPA), and also will briefly critique the general concept of “auto-fill forms” and the need to treat them much more like a pseudo-user-login forms than some sort of marketing growth hack, which got Facebook into this mess in the first place.</p><p>Without further adieu…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kUS480LXK3bPf6iweyx9sw.jpeg" /></figure><h3>Facebook user flow prior to accidental “mkt_tok” one-token-user-authentication vulnerability</h3><p>While disclosing this to Facebook, I showed them a video of me navigating from a Facebook Developer Newsletter to an Instagram + Adobe Spark signup webpage @ <a href="https://go.fb.com/sparkAR-instagram.html">https://go.fb.com/sparkAR-instagram.html</a>, and then how the form on that page was using Marketo auto-fill form technology via one-token-user-authentication. There are screenshots of that process below.</p><p>After disclosure, Facebook removed the form I was testing on and redirected that page to a new secure form @ <a href="https://www.facebook.com/arp/ig/beta">https://www.facebook.com/arp/ig/beta</a>, but dozens of other legacy forms with the original vulnerability exists across Facebook domains, and Facebook continues to send out the “mkt_tok” user token-appended URLs on a daily basis via multiple Facebook divisions/programs.</p><p>Globally, any legacy form with Marketo auto-fill technology that uses one-token-user-authentication via “mkt_tok” and is “properly” setup via Marketo’s CNAME DNS data mapping, can be turned into a user data scraping machine with minimal effort to pull data from a client Marketo account. By merely copying/pasting/replacing the “mkt_tok” text string parameter in any URL hosting a legacy Marketo form with “auto-fill” turned on, and by using a new cookie-free session to make each page request, Marketo will append unique user data into a form via auto-fill (and also pass that user data via HTMl/the data layer, depending on the page setup). For custom forms, any unique form fields will also leak unique data. Furthermore, the unsubscribe pages and other email preference pages built by Marketo and hosted on client subdomains are ripe for additional data leaking via mkt_tok one-token-user-authentication, but fortunately, Facebook has a custom setup that requires multiple tokens to access these pages.</p><p>Facebook’s user data exfiltration vulnerability only exists with auto-fill Marketo forms on their pages, but this is still a Facebook user data exfiltration vulnerability and far too simple for an attacker to execute. This is what a vulnerable URL looks like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uR0G59xKzQl__X_y5hFgEw.jpeg" /></figure><p>Here’s a simplistic technical map of this data flow from DNSdumpster.com showing how the insecure <a href="http://content.fb.com/">http://content.fb.com</a> subdomain is CNAME-mapped to Adobe’s Marketo to share user data between the two companies via this non-SSL subdomain.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4VVAj_UwtcRCwYqqaki3rw.jpeg" /></figure><p>Here’s a screenshot walkthrough of a Facebook Developer newsletter email that contained the “mkt_tok” one-token-user-authentication vulnerability, and the front-end user experience as data is leaked.</p><h3>Step One: Facebook Developer Gets an Email from Facebook Promoting Facebook Developer News</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d5ibfnGlSpYN3sCTHFZBzA.jpeg" /></figure><h3>Step Two: After Clicking the Email, the User is Sent Through a URL Redirect on <a href="http://content.fb.com/">http://content.fb.com</a>,</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u6ObD0uqvnWmEhaOYIxG2Q.jpeg" /><figcaption>Extremely simple javascript code hosted on <a href="http://content.fb.com/">http://content.fb.com</a> and generated via Adobe Marketo to facilitate the URL redirect from a Facebook Developer newsletter, and also append the “mkt_tok” one-token-user-authentication.</figcaption></figure><p>This insecure redirect is hosted on content.fb.com in dynamically generated (but permanent) pages and uses a javascript snippet to append a “mkt_tok” one-token-user-authentication URL variable to the final resolved URL.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NU-uXb_0ivP47tOzTx4K-g.jpeg" /></figure><h3>Step Three: Facebook User is Redirected to the Final Resolved URL with “mkt_tok” Appended as a URL Parameter, Open and Exposed — a Growth Hack Feature and a User Data Vulnerability</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MbN7B5DZNPvBUAsTuqIqiQ.jpeg" /></figure><h3>Step Four: On a “Properly Setup” Page with Adobe Marketo Auto-Fill Form Technology, the User Data is Exposed by Only Appending the “mkt_tok” Parameter to the URL</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gjTsjaoZId4RdWBuydDSqQ.jpeg" /></figure><h3>Step Four Bonus: Facebook Users Easily Exposed Their “mkt_tok” URLs via Native Facebook Sharing Tools</h3><p>Unfortunately, many Facebook users exposed themselves and their own “mkt_tok” user parameter, including the VIP marketer pictured at the top of this post with Facebook CEO Mark Zuckerberg.</p><p>This is how users exposed their own “mkt_tok” token by promoting Facebook news:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Rb0WokQ39l5YbJlxvyHvAQ.jpeg" /></figure><h3>Facebook’s Adobe Marketo problem and the “mkt_tok” one-token-user-authentication user data vulnerability comes from <a href="http://content.fb.com/">http://content.fb.com</a></h3><p>Facebook’s user data exfiltration vulnerability impacts Facebook Workplace users, Facebook Marketing IQ program users, Facebook Developers, and Facebook for Business users.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OX9T_XAmwokRWiOnUg4WaQ.jpeg" /></figure><p>Facebook’s vulnerability exists due to their implementation of software owned by Adobe called Marketo — <strong>this Marketo email software is installed on the Facebook-controlled </strong><a href="http://content.fb.com/"><strong>http://content.fb.com</strong></a><strong> subdomain via CNAME DNS record mapping and the Marketo software is then used by internal Facebook marketing staff to promote Facebook programs and tools.</strong></p><p>The insecure subdomain used in this data flow is ripe for attack, both locally and at a network level, but more importantly is an avenue for user data exfiltration.</p><p>From my perspective, I don’t totally understand all of the vulnerabilities with this subdomain being non-SSL, beyond the issues outlined in this post. But I look forward to more external developers viewing and analyzing the <a href="http://content.fb.com/">http://content.fb.com</a> subdomain and the redirect javascript to provide additional context. I won’t, however, share one of the redirect URls, you’ll have to find one yourself or signup for Facebook newsletter emails.</p><h3>Deployment of Adobe Marketo in the Facebook Workplace New User Onboarding Emails</h3><p>Facebook uses Adobe’s Marketo email software for a wide variety of their programs, but many people would be surprised to find that every new user of Facebook Workplace will receive emails containing their private“mkt_tok” token.</p><p>These insecure emails are part of an “onboarding email series” to get new users familiar with Facebook Workplace. At least one email in this Facebook Workplace new user onboarding links users to Facebook’s <strong>TransformYourWorkplace.com</strong>, which also passes those visitors “mkt_tok” user data to Google Analytics.</p><p>One email in the onboarding flow is un-ironically titled, “Keep your data safe with Workplace” and the link in the email appends the “mkt_tok” user token:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bm-QnNOYp9dTVd2cNTHRCA.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oE5B6FTo9is6rL_54wvZfw.jpeg" /></figure><p><strong>Not-so Safety Check</strong></p><p>Additionally, Workplace by Facebook has been onboarding “Early Access” to their “Safety Check” feature for workplace emergencies, and built a webpage and form for this using Adobe’s Marketo software, a form that has the user data exfiltration vulnerability. This form also includes a custom field that can be scraped (if a user submitted that specific form and also had their “mkt_tok” leaked) that asks users the open-ended question, “Anything else we should know?”</p><h3>Want to signup for email updates via Facebook’s Newsletters or Facebook Workplace?</h3><p>As of publishing, each of these Facebook newsletter signups could lead to you receiving an email with an insecure “mkt_tok” token. You shouldn’t share any link containing this URL parameter publicly.</p><blockquote><a href="https://www.facebook.com/business/insights/newsletter-sign-up">The Facebook Business Insights Newsletter</a></blockquote><blockquote><a href="https://www.facebook.com/business/insights/newsletter-sign-up">The Facebook Marketer IQ Newsletter</a></blockquote><blockquote><a href="https://developers.facebook.com/">Facebook Developer Updates</a> (You’ll need to click around, these signups are across a bunch of programs)</blockquote><blockquote><a href="https://workplace.facebook.com/">The Facebook Workplace New User Signup</a></blockquote><h3>Facebook Confirmed that Adobe Marketo’s Email Links are Unsafe for Users</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/980/1*1qB4byX-JB0qlWF19A5HVA.jpeg" /></figure><p>Facebook was initially informed about this user data vulnerability on October 29th, 2018, but my initial email didn’t properly explain the form scraping vulnerability, and their initial response on November 1st, 2018 rejected any security concerns and stating, “This is intentional behavior in our product” and “While it’s true that these links, when inadvertently shared the way that you’ve shown they can be, may lead to unintended changing of a user’s email preferences, we do not encourage users to share these links or to make them publicly available. As such, we do not consider it a security vulnerability, but we do have controls in place to monitor and mitigate abuse.”</p><p><em>((NOTE: Facebook uses Cloudflare on </em><a href="http://content.fb.com/"><em>http://content.fb.com</em></a><em> to protect against malicious script injections on their insecure subdomain, but Cloudflare can’t do anything to protect against a “software feature” that facilitates user data scraping with a known “mkt_tok” variable. Cloudflare can protect against all kinds of URL injections and automated scripts, but I saw almost no Cloudflare errors during my tests because it’s considered “intended user behavior” to auto-append the data into the forms))</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/976/1*JqRZMo-g_3z9xyHSj035hQ.jpeg" /></figure><p><strong>After Facebook’s initial rejection of security concerns, </strong>I recorded a video (the screenshots above in this post are from that video), and further explained how the one-token-user-authentication “mkt_tok” URL variables were being appended to *all* links being sent out in these newsletters, and also how the auto-fill forms they were using facilitated the scraping of user data using any exposed user tokens. After a bit of back and forth, on November 18th, 2018, Facebook asked for “concise reproduction instructions/steps for how an attacker would exploit this token after the victim unwittingly shares it?”</p><p>I responded to Facebook on November 19th with additional steps on how an attacker could append a user “mkt_tok” one-token-user-authentication to a URL on Facebook that was hosting an Adobe Marketo email form with auto-fill technology to access and scrape the user data. On November 29th, 2018, Facebook security sent a rambling response that †ruly must be read to be believed:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/986/1*sVs31BqkV5IRaQHkIYYo9Q.jpeg" /></figure><p>After several additional messages, Facebook responded on December 3rd, 2018 writing, “At this time sharing those links is not recommended. While not ideal those links are generated by Marketo, which is not a Facebook product.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/990/1*8EzKse0bqqbn8DR4YpPEhw.jpeg" /></figure><h3>Facebook doesn’t consider one-token-user-authentication secure, and will admit as much, but hasn’t changed their practices…. so what next?</h3><p><strong>This data exfiltration vulnerability appears to only impact a small number of Facebook users</strong> — albeit mostly VIP users and core platform developers and marketers.</p><p>One of the most alarming parts of this disclosure is that even though it was mentioned to them many times, they have taken no efforts to remove the ability to search for this exposed token via Facebook’s native search. As of publishing time, <a href="https://www.facebook.com/search/str/%22mkt_tok%22/keywords_blended_posts?se2856=1&amp;epa=FILTERS&amp;filters=eyJycF9hdXRob3IiOiJ7XCJuYW1lXCI6XCJtZXJnZWRfcHVibGljX3Bvc3RzXCIsXCJhcmdzXCI6XCJcIn0ifQ%3D%3D">you can still go on Facebook and search for “mkt_tok”</a> (*and other insecure one-token-user-authentication URL variables from other software companies*) and view not only the Facebook users exposing themselves via Facebook’s programs but also users exposing themselves from links sent by non-Facebook Adobe Marketo clients.</p><p>Facebook made a few minor changes as I was reporting these issues to them, but the issue persists and user tokens continue to get exposed. This is also only the tip of the iceberg for one-token-user-authentication URL variables and the need for software vendors to move away from this insecure authentication protocol. Unfortunately, other insecure URL variables are also being shared on Facebook daily without being removed. The question remains, who is responsible for protecting a user from themselves when a URL parameter can be used to access their private information?</p><h3>Adobe was Notified of Marketo’s “mkt_tok” User Data Vulnerability on September 25th, 2018, Over a Month Before Completing the Purchase of Marketo on October 31st, 2018</h3><p>Adobe’s security team is typically very responsive via email and they obviously triage some security problems within hours of an initial report. As was typical, in the screenshot below you will see their team responding to my initial email to them on September 25th, 2018 about the Marketo “mkt_tok” user token vulnerability — I sent this email to them after seeing news reports that they were in the process of acquiring Marketo. They opened a ticket (PSIRT -8927) about the report on September 25th, but went silent on it. Adobe completed the acquisition of Marketo on October 31st and still hasn’t followed up.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SuiegEGIRd7CBmuxHkZ6gA.jpeg" /></figure><h3>Adobe *had* user data problems similar to Marketo, and were alerted to a similar one-token-user-authentication user data vulnerability via Adobe’s “pkey” URL variable on April 2, 2018</h3><p>Way back on April 2, 2018, I emailed Adobe and alerted them to a vulnerability impacting their “pkey” one-token-user-authentication URL variable and client subdomains that were caching user sessions via this URL variable and putting Adobe end-user data at risk (mostly exposing user emails and certain documents that were being emailed to users). The initial organization I flagged for them which had user data exposed was their client the Boy Scouts of America. They responded to my initial email in under 90 minutes and from my limited-view of their response efforts, they moved pretty swiftly from a couple fronts to fix the problems. Since then, and prior to their completed acquisition of Marketo, it appears as though they removed all one-token-user-authentication from their software stacks.</p><p>Here’s a screenshot of my initial email and Adobe’s PSIRT case #8018 — this was from April 2, 2018:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rrxwexBK8Z2dLDQXvV5HnQ.jpeg" /></figure><p>For the last few months, even though someone could find remnants of problems across Adobe client installs, (google the phrase: “inurl:pkey”), it would appear that every one-token-user-authentication problem of Adobe’s seems to have been fixed by denying the requests (aka breaking the URLs) — this basically broke lots of links and may have created problems for clients — but it also prevented any data exposure and was probably the correct decision. This was a bold choice from Adobe that I was surprised they never publicized, but I suspect they didn’t want to publicly admit to the data exposure via URL token caching since it was probably resolved before the GDPR deadline.</p><p>But now, <a href="https://news.adobe.com/press-release/corporate/adobe-completes-acquisition-marketo">Adobe owns Marketo as of October 31st, 2018</a>, and almost assuredly is planning the same process to make Marketo secure that they deployed across Adobe earlier in 2018, but Adobe has neither publicly disclosed the ongoing security problems with their new acquisition Marketo, privately reported the ongoing security problems to their Marketo clients or taken the bold-yet-necessary steps to just break all existing Marketo client URLs that attempt to authenticate and access user data via the one-token-user-authentication “mkt_tok” URL variable.</p><p><strong>Adobe has experience fixing security problems from one-token-user-authentication via their “pkey” URL variable — when will they take steps to fix their “mkt_tok” URL variable?</strong></p><h3>The California Consumer Privacy Protection Act (CCPA) Includes Protections for User Tokens</h3><p>The California Consumer Privacy Protection Act (CCPA) signed into law and going into effect in January 2020 <a href="https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=201720180SB1121">includes this paragraph</a> when explaining the scope of the law (and why I’d argue CCPA includes protections for “mkt_tok” from being shared/used without consent):</p><blockquote>“(x) ‘Unique identifier’ or ‘Unique personal identifier’ means a persistent identifier that can be used to recognize a consumer, a family, or a device that is linked to a consumer or family, over time and across different services, including, but not limited to, a device identifier; an Internet Protocol address; cookies, beacons, pixel tags, mobile ad identifiers, or similar technology; customer number, unique pseudonym, or user alias; telephone numbers, or other forms of persistent or probabilistic identifiers that can be used to identify a particular consumer or device. “</blockquote><p>It shouldn’t come as a surprise that legal opinions will eventually determine that CCPA includes protections for “mkt_tok” and the type of URL parameter/cookie with one-token-user-authentication privileges that can not only be used to track a user but also access their private information.</p><h3>GDPR Already Includes Protection for “mkt_tok” and Other User Tokens</h3><p>The General Data Protection Regulation framework implemented across Europe on May 25, 2018, <a href="https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32016R0679">includes this paragraph</a> when explaining the scope of the law, a clear reference to user tokens like “mkt_tok”:</p><blockquote>“Natural persons may be associated with online identifiers provided by their devices, applications, tools and protocols, such as internet protocol addresses, <strong>cookie identifiers</strong> or other identifiers such as radio frequency identification tags. <strong>This may leave traces which, in particular when combined with unique identifiers and other information received by the servers, may be used to create profiles of the natural persons and identify them</strong>.”</blockquote><p>So the question remains for Adobe and Facebook, have European Authorities been notified of their “mkt_tok” user data exfiltration vulnerabilities and ongoing data leaks?</p><h3>Auto-fill forms are inherently insecure — it’s very uncommon to see secure setups</h3><p>During the research for this post to better understand the types of user data authentication being used by “enterprise” marketing apps, it became abundantly clear that there aren’t many ways to have an auto-fill form without creating unique user data vulnerabilities. The reality is that anytime a plain-text email address or other user PII is “pasted” into a webpage, that data still needs to remain encrypted and only partially visible (aka partially hashed emails). Otherwise, one bad development decision can turn into a massive data supply-chain leak if your marketing team starts to fire 3rd party advertising and analytics pixels for those website visitors, which passes the full URL to other 3rd-party companies (<em>and sometimes text/fields on the page, see Adroll’s “Data Shotgun” approach</em>). As of publishing, there are unfortunately dozens of zero-day data exfiltration vulnerabilities using one-token-user-authentication across the marketing software ecosystem.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/980/1*1qB4byX-JB0qlWF19A5HVA.jpeg" /></figure><p>There are also dozens of enterprise marketing companies that create user unsubscribe/preference pages that are insecure. Facebook itself argued that even though the “mkt_tok” user variable could let an attacker access and change a Facebook user’s email preferences, that wasn’t a security vulnerability — and Facebook even had the gall to cite an FTC advisory opinion about un-subscribe links and CAN-SPAM compliance. :D</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jDbHOogpbpr19R4S4Wo9Sg.jpeg" /></figure><p>If Facebook had been paying attention, industry leaders like MailChimp have been hashing user emails on email unsubscribe/preference pages for years — aka, following the FTC opinion but also not putting user data at risk of exfiltration. MailChimp is, unfortunately, the exception and there are plenty of insecure enterprise email tools….but other companies insecure existence doesn’t mean their data architecture should be modeled by massive companies like Facebook.</p><h3>Conclusion — Facebook and Adobe Need to Answer Questions and Remove User Data Exfiltration Vulnerabilities from “mkt_tok”</h3><p>The information contained in this post was previously disclosed to Facebook, Adobe, and Marketo across several long email chains. Adobe appears to have taken little action to resolve or remedy their vulnerabilities since finishing the acquisition of Marketo on October 31, 2018, and took some steps that made it seem to indicate they had no intention of publicly warning their clients or reporting these vulnerabilities as data controllers under GDPR. Facebook has made a few minor changes, but continues to send out emails regularly with the “mkt_tok” user token and continues to host Marketo forms with auto-fill technology that is vulnerable to user scraping.</p><p>One of the most interesting aspects of this research has been the clear admissions from Facebook security staff that they didn’t think Adobe’s Marketo email software was secure for users. The facts are pretty clear about how these tokens are exposed daily, and how they can be used to access user data, yet it seems like either Facebook is waiting for Adobe to fix the problem, or Adobe is waiting to deny the insecure “mkt_tok” one-token-user-authentication user data requests until after they built a new data architecture.</p><p>If both Adobe and Facebook are basically partnering with a custom deployment of Marketo, and exposing user data across the global supply chain, who has exposure as the data controller? Both of them? Only the originator of the user agreement, aka Facebook?</p><p>Significant questions remain about the legal exposure of enterprise companies who deploy insecure software from other companies, and their unique data controller requirements under existing user privacy frameworks.</p><h3>Questions that need answers :</h3><p>When Facebook sent an email newsletter that linked Facebook Developers to the European Tech Publisher TheNextWeb (<a href="https://thenextweb.coxm/syndication/2018/02/11/can-train-ai-convert-design-mockups-html-css/">this blog post of theirs</a>), while also appending the “mkt_tok” URL variable to the final resolved URL which was passed to TNW and dozens of their advertising partners, would Facebook be required to submit information about this to European regulators or request deletion?</p><p>Should Facebook and Google let users search for URL parameters? Should it be so easy to go on Facebook and search “mkt_tok” to find exposed users?</p><p>Should all software companies be required to organize and publish their one-token-user-authentication URL parameters so other companies can protect themselves from inadvertently ingesting or sharing those user tokens?</p><p>Should Facebook, Google, Cloudflare and other major service providers publish lists of URL parameters and cookie variables (and known details about them) to help publishers and other companies protect themselves from inadvertently ingesting or sharing those user tokens?</p><p>Should all massive companies map the URL parameters and user tokens created across their business and subsidiaries? Should companies track and create user data maps to determine data exposure and leaks along a data supply chain and along the user consent chain?</p><h3>Questions? Concerns? Revisions?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/620/1*UgcSSG61qgGzaul6BYZrnw.jpeg" /></figure><p>Tweet me @<a href="https://twitter.com/thezedwards">thezedwards</a>, @ zach@victorymedium.com, or a spam Linkedin connection request <a href="https://www.linkedin.com/in/zedwards/">here</a>… and while you’re at it, want to check out <a href="https://medium.com/@thezedwards/the-cookie-exchange-how-gdpr-created-a-new-market-for-pseudonymized-cookies-592cbde9b44b">my proposal for a new digital Cookie Exchange</a> using the Shapley Value formula and ETF baskets traded on a public market to automate click arbitrage? <a href="https://medium.com/@thezedwards/the-cookie-exchange-how-gdpr-created-a-new-market-for-pseudonymized-cookies-592cbde9b44b">Click here to wonk out</a>. Thanks for any feedback!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d8435a259b0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Cookie Exchange — How GDPR Created a New Market for Pseudonymized Cookies]]></title>
            <link>https://medium.com/@thezedwards/the-cookie-exchange-how-gdpr-created-a-new-market-for-pseudonymized-cookies-592cbde9b44b?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/592cbde9b44b</guid>
            <category><![CDATA[gdpr]]></category>
            <category><![CDATA[cookies]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[advertising]]></category>
            <category><![CDATA[stock-market]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Fri, 25 May 2018 23:51:10 GMT</pubDate>
            <atom:updated>2018-06-06T18:46:09.230Z</atom:updated>
            <content:encoded><![CDATA[<p><em>6/6/18 Update — </em><a href="https://www.youtube.com/watch?v=NKkXCHBW2d8"><em>Click here to watch a summary video of this proposal </em></a><em>— the video is 45 minutes long but attempts to explain some of the more complicated details. Thanks for watching and reading and all the feedback!</em></p><p>GDPR goes into effect today, May 25th, 2018, and creates a massive shift in global expectations for user privacy.</p><p>Users have made it clear that the state of advertising and tracking is unacceptable, publishers are going broke, and the market continues to swing more and more out of balance with Google, Facebook, Amazon and Apple making record profits. And until today, companies have not faced any real financial incentives to pseudonymize data.</p><p>Though many perceive GDPR as effectively outlawing certain data markets, the implementation of GDPR also poses a unique opportunity to remake markets and monetize data — without compromising user privacy and while promoting competition and undermining existing monopolies.</p><p><strong><em>Before you read any further, I’d strongly recommend watching this short video, below, from Bloomberg News about ETF Markets and how they stabilize prices and create new value. This will provide a good primer on the proposed Cookie Exchange concept and how an ETF Market built on the arbitrage of pseudo-cookies can be used to automate online advertising targeting while stabilizing costs.</em></strong></p><p><a href="https://www.bloomberg.com/news/videos/2016-03-07/what-is-an-etf">What Is an ETF?</a></p><h3><a href="https://www.bloomberg.com/news/videos/2016-03-07/what-is-an-etf">“There are now over 6,000 ETFs on 60 exchanges and ETFs exist for everything from corporate bonds to gold bars to oil…” — Click here to watch the 2-minute video on Bloomberg.com.</a></h3><h4>The Genesis of the Cookie Exchange</h4><p>In my early 20s I was a professional poker player living in Vegas. I’ve always loved proposition bets — basically any bet made outside a traditional card game or casino game. I left poker over a decade ago, but the data-driven approach to game theory has served me well during my career in digital.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tI8qdFTV2xa6j472D8j9EQ.jpeg" /></figure><p>I had a bit of a eureka moment about a year ago when I thought to offer a fellow marketer a bet:</p><h4>“I’’ll bet you $50 with Texas-sized odds @ 3:1 that you can’t beat $1.75 cost-per-click @ $45 cost-per-customer-acquisition for the Facebook Audience ‘Newly Moved’ and a budget of $1k.”</h4><p>At that moment, I realized a potential market existed for the exchange of cookies. And we could create that market with incentives, similar to emerging market ETFs, to stabilize ad-click costs, and manage global advertising targeting through automated algorithms. This pseudo-arbitrage would reduce costs, promote data sharing, and encourage cooperation through a variation of the Shapley Value formula. Delivery of ads would be optimized with variations of the <a href="https://en.wikipedia.org/wiki/Multi-armed_bandit">multi-armed bandit algorithms</a> and the <a href="https://www.microsoft.com/en-us/research/publication/web-scale-bayesian-click-through-rate-prediction-for-sponsored-search-advertising-in-microsofts-bing-search-engine/?from=http%3A%2F%2Fresearch.microsoft.com%2Fpubs%2F122779%2Fadpredictor%2520icml%25202010%2520-%2520final.pdf%2520%28http%3A%2F%2Fresearch.microsoft.com%2Fpubs%2F122779%2Fadpredictor%2520icml%25202010%2520-%2520final.pdf%29">Bayesian algorithms that Microsoft</a> and other companies use to historically optimize click-through-rates for ads. The tech is here, just being applied for a different market.</p><p>This new Cookie Exchange would not only provide massive new audiences for brands and companies by opening walled-gardens into public exchanges, but by also harmonizing GDPR’s pseudonymization and right to be forgotten principles with the digital world’s demand for data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5UuZPPcjPs_zJCqhJpmwLg.jpeg" /></figure><h3>What is “Pseudonymization” and Why Does it Matter?</h3><p>GDPR effectively requires companies to either anonymize user data or pseudonymize user data, if they intend to use it beyond the originally authorized use (<em>See GDPR, Article 6, Section 4</em>). <a href="https://en.wikipedia.org/wiki/Pseudonymization">Wikipedia describes pseudonymization</a> and it’s slight difference from completely anonymous data:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hw9PhPLAWcKLJvpNzvP-BQ.jpeg" /></figure><blockquote><strong>Pseudonymization</strong> is a <a href="https://en.wikipedia.org/wiki/Data_management">data management</a> and <a href="https://en.wikipedia.org/wiki/De-identification">de-identification</a> procedure by which <a href="https://en.wikipedia.org/wiki/Personally_identifiable_information">personally identifiable information</a> fields within a <a href="https://en.wikipedia.org/wiki/Data">data</a> record are replaced by one or more artificial identifiers, or <a href="https://en.wikipedia.org/wiki/Pseudonym">pseudonyms</a>. A single pseudonym for each replaced field or collection of replaced fields makes the data record less identifiable while remaining suitable for <a href="https://en.wikipedia.org/wiki/Data_analysis">data analysis</a> and <a href="https://en.wikipedia.org/wiki/Data_processing">data processing</a>.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uqL5awzEKmQwcvTKcotrGw.png" /></figure><p>The major difference between pseudonymized data and any data that is tied directly to a user’s personally identifiable information, is the fact that the latter is permanent.</p><p>Pseudonymized data allows a user to essentially tell advertising networks, “I want to see ads on X topic or product” — and then when the usefulness of that has expired, the user could delete their cookie and start over with a fresh profile. The “pseudonymizedID” associated with every cookie could be deleted and then a totally new pseudonymizedID could be acquired by a user to start over with the data they are sending to advertisers.</p><p>Some core social networks like Facebook and Google would provide significant amounts of audience data for this initial Cookie Exchange (and the Shapley Value Formula accounts for that, more below). In order to have a stable inventory of data and ads, organizations would be able to “back-fill” their own targeting data into a pseudonymized cookie, but only putting that data into the device and browser a user was currently using — essentially every device generates unique pseudonymized cookies and they are constantly expiring both from user action, but also due to jurisdiction requirements.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fr6TZ-Bgf_YkfkNBsZhENw.jpeg" /></figure><p>There are many strategic benefits for using pseudonymized data, with market automation being paramount among them. Data that contains PII literally can’t be traded in the EU en-masse since GDPR was passed — so the only way to actually service a stable ads market is to support pseudonymized cookies and efforts to let users clear them at will and networks to refill them and share them for ads without compromising the user’s privacy.</p><p>The Public “Cookie Exchanges” via <a href="https://en.wikipedia.org/wiki/Exchange-traded_fund">ETF Markets</a> (an arbitrage market that reduces and normalizes costs) could be created to properly reward the creation of this pseudonymized cookie data.</p><h3><strong>Digital Cookies Used for User Tracking Have Value</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IIkhy061285CDcWu9_EMNg.jpeg" /></figure><p>One of the most important parts of the new Cookie Exchange is <strong><em>acknowledging the existing value</em></strong> in all of the cookies (user data) that websites, apps and other technology are currently giving away to a couple companies in Silicon Valley.</p><p>The vast majority of websites act as share-croppers and receive $0 revenue from the data they provide the current dark ads market. And even the biggest publishers don’t know how to track their data leaks, or how to ask for their money.</p><p>But that doesn’t need to be the case. The Shapley Value Formula, currently being used for Google’s newest data-driven attribution models, if structured slightly different, could create <strong>a new cooperative publisher environment where dozens of companies, publishers and content creators could receive money from every advertising click.</strong></p><p>The Shapley Value is a game theory concept driven by the inherent value of cooperation. Under the Shapley Value, collaborators with different needs and different contributions maximize their return by ensuring that each actor gains as much or more as they would have from acting independently.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tVp6jwyMrqtHXPQnB1zGdA.jpeg" /><figcaption>Massol and Pierru’s variation of the Shapley Value Formula from “<em>The Shapley Value as a tractable solution to share costs among a large number of substitutable customers. A Gas Transmission example.</em>” (<a href="http://www.iaee.org/en/publications/proceedingsabstractdoc.aspx?id=1224">View abstract of the paper — warning Word DOC</a> from iaee.org — <a href="http://www.iaee.org/en/publications/proceedingssearch.aspx?conference=Briding+Energy+Supply+and+Demand%3A+Logistics%2C+Competition+and+Environment">purchase the paper here for $15</a>.)</figcaption></figure><p>The Cookie Exchange version of the Shapley Value Formula was originally developed to share costs of natural gas transmission lines across diverse customers, and was published in a paper by Olivier Massol and Axel Pierru in 2008.</p><p><em>(Note on the Shapley Value Formula: cooperation algorithms and game theory are discussed in </em><a href="https://www.researchgate.net/publication/260497493_A_Nash_Approach_to_Planning_Merchant_Transmission_for_Renewable_Resource_Integration"><em>a wide variety of oil and gas papers</em></a><em> and by other academics — they </em><a href="https://www.tandfonline.com/doi/abs/10.1080/00137910701854602"><em>should be weighed</em></a><em> against Massol and Pierru’s formula, and it would seem certain that even MP’s base formula would see many changes before being deployed into the Cookie Exchange. Latency of the formula execution will be one of the key differentiators — the Massol and Pierru formula can split cooperatively across 100 segments (publishers) in a thousandth of a second. Massol has published </em><a href="https://www.city.ac.uk/__data/assets/pdf_file/0005/90059/1103_massol.pdf"><em>another paper on Shapley optimization</em></a><em> , but in general, the research into cooperative markets is too vast to get into deeply in this article.)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q_ZKnMNq7pNUOSFxHm3tdQ.jpeg" /></figure><h3>Why Google Uses the Shapley Value Formula for Attribution and Why the World should Use it to Cooperate on the Internet</h3><p>Google uses the Shapley Value Formula to determine cooperation between sources of a conversion and attribute revenue proportionally to each of them. <a href="https://www.pathinteractive.com/blog/digital-marketing/the-science-behind-data-driven-attribution/#.WwZA_tMvwk8">Steven Sonnes from Path Interactive described</a> Google’s strategy as a “cooperative game theory”:</p><blockquote>In 2013 Google unveiled the Data-Driven attribution model for Google Analytics Premium, and in 2014 they released it in AdWords. …Data-Driven Attribution uses an algorithm based on a concept from cooperative game theory called the <a href="https://en.wikipedia.org/wiki/Shapley_value">Shapley Value</a>. The Shapley Value was developed by the economics Nobel Laureate Lloyd S. Shapley as an approach to fairly distributing the output of a team among the constituent team members. In the case of Data-Driven Attribution, the “team” being analyzed has marketing touchpoints (e.g., Organic Search, Display, and Email) as “team members,” and the “output” of the team is conversions. The Data-Driven Attribution algorithm computes the counterfactual gains of each marketing touchpoint — that is, it compares the conversion probability of similar users who were exposed to these touchpoints, to the probability when one of the touchpoints does not occur in the path.</blockquote><p>In plain English, the Shapley Value rewards market players who collaborate with each other, to share value, when such cooperation ensures the same or greater gains than if they acted independently.</p><h4>Google DDA Attribution (Shapley Value) Accurately Gives Credit for Past Touch Points</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iyG_C55lvgRdd4fc7Wvubw.jpeg" /></figure><p>Sabrina Garufi, Agency Development Manager at Google, gave a Google DDA Attribution presentation for Click Consult’s Benchmark Search Conference 2017 — <a href="https://www.slideshare.net/clickconsult/measuring-complex-customer-journeys">view the full presentation here</a> or embedded below.</p><p>Slides 19 and 20 give a simple high-level view of how purchase conversions occur from advertising touch points over days and weeks. Google DDA Attribution, using the Shapley Value to encourage cooperative and balanced attribution, splits one sales conversion into dozens of pieces so that each advertising touch point is appropriately credited.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.slideshare.net%2Fslideshow%2Fembed_code%2Fkey%2FkF98esadt40SLT&amp;url=https%3A%2F%2Fwww.slideshare.net%2Fclickconsult%2Fmeasuring-complex-customer-journeys&amp;image=https%3A%2F%2Fcdn.slidesharecdn.com%2Fss_thumbnails%2Fsabrinagarufi-171002080819-thumbnail-4.jpg%3Fcb%3D1506932515&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=slideshare" width="600" height="500" frameborder="0" scrolling="no"><a href="https://medium.com/media/9c6c119b6f0a8ad67db69c38cdac9dcc/href">https://medium.com/media/9c6c119b6f0a8ad67db69c38cdac9dcc/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/216/1*VIN4ddpVZ3Xx-vp29O3iXA.jpeg" /></figure><h4>If Google uses the Shapley Value to credit “traffic sources” (but not give them any revenue) why can’t we create a new market that provides monetary value for the “conversion assists” on a user’s pseudo-cookie?</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OxoVVR6ATNfmzcShz2AkdA.jpeg" /></figure><h3>Security of the Cookie Exchange from Current “Same-Origin” Tech Standards — Pseudonymized Cookie Data Passing Between Domains and Businesses Securely via “Cross-Origin Resource Sharing” (CORS) via the “withCredentials” Property</h3><p>Dealing with internet/web cookies is complicated, and many parts of this Cookie Exchange may require some tweaks to schema standards. Fortunately, the core tech standards already exist for both Cookie Exchanges and for the pseudonymized cookies to pass between separate websites and unique businesses.</p><p><a href="https://en.wikipedia.org/wiki/Same-origin_policy">The Same-Origin Wikipedia article</a> gives a good explanation about this type of internet security, but if you’ve ever heard of “<a href="https://en.wikipedia.org/wiki/Cross-site_scripting">Cross-Site Scripting</a>” (XSS) or wondered how banks and secure websites keep your information private — they do it through this core security built into all internet connections and devices. Here’s how Wikipedia describes how one website prevents you from being attacked by another:</p><blockquote>In <a href="https://en.wikipedia.org/wiki/Computing">computing</a>, the same-origin policy is an important concept in the <a href="https://en.wikipedia.org/wiki/Web_application">web application</a> <a href="https://en.wikipedia.org/wiki/Computer_security_model">security model</a>. Under the policy, a <a href="https://en.wikipedia.org/wiki/Web_browser_engine">web browser</a> permits scripts contained in a first web page to access data in a second web page, but only if both web pages have the same <em>origin</em>. An origin is defined as a combination of <a href="https://en.wikipedia.org/wiki/Uniform_Resource_Identifier">URI scheme</a>, <a href="https://en.wikipedia.org/wiki/Hostname">host name</a>, and <a href="https://en.wikipedia.org/wiki/Port_(computer_networking)">port number</a>. This policy prevents a malicious script on one page from obtaining access to sensitive data on another web page through that page’s <a href="https://en.wikipedia.org/wiki/Document_Object_Model">Document Object Model</a>.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cYI52Rb-fWjSFiQPoCF9pQ.png" /><figcaption>CORS Flow Chart from <a href="https://www.html5rocks.com/en/tutorials/cors/">HTML5Rocks.com</a></figcaption></figure><p>This core Same-Origin Policy has evolved for some “wiggle room” ($$) to allow for “<a href="https://en.wikipedia.org/wiki/Cross-origin_resource_sharing">Cross-Origin Resource Sharing</a>” (CORS) — which allows for certain types of data (cookie files and some other types of files) to be passed between websites and other systems.</p><h4>It’s clear that the schema for the Cookie Exchange would need to be debated for consensus among publishers, but there are no technical limitations to pass pseudonymized data between domains via a cookie text file.</h4><p>More specifically, CORS would use the “withCredentials” property to send and set cookies between domains. HTML5Rocks’ <a href="https://www.html5rocks.com/profiles/#monsurhossain">Monsur Hossain</a> did a <a href="https://www.html5rocks.com/en/tutorials/cors/">fantastic job with the article on CORS</a> and the deployment of “withCredentials” to set cookies. To the left on this page is the flow chart from Hossain that shows how one domain asks another domain for cookie information and then has it passed securely:</p><p>There are a bunch of other great resources on CORS, including from <a href="https://www.nczonline.net/blog/2010/05/25/cross-domain-ajax-with-cross-origin-resource-sharing/">NCZOnline</a>, <a href="https://www.html5rocks.com/en/tutorials/cors/">HTML5Rocks</a>, <a href="http://promincproductions.com/blog/cross-domain-ajax-request-cookies-cors/">Brian Prom</a>, <a href="https://stackoverflow.com/questions/36365409/setting-cookies-with-cors-requests">Stack Overflow</a>, <a href="https://quickleft.com/blog/cookies-with-my-cors/">Cognizant’s Quick Left</a>, and <a href="https://developer.mozilla.org/en-US/docs/Web/HTTP/CORS">Mozilla</a>.</p><h4>CORS is part of the magic sauce behind a Cookie Exchange actually passing data between websites…but one of the most exciting things about the Cookie Exchange is that websites aren’t the only thing that can append data to a cookie…</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3WCv5aeWFElg1pU-BrG00w.jpeg" /></figure><h3>How the Cookie Exchange ETF Market Would Trade Cookie Indexes and Baskets</h3><p><em>If you haven’t already, make sure to check out</em><a href="https://www.bloomberg.com/news/videos/2016-03-07/what-is-an-etf"><em> this video from Bloomberg News about ETF markets</em></a><em>.</em></p><h4>How would a banker and anyone participating in an ETF Market stabilize prices and automate targeting?</h4><p>In a dramatically simplified example, basically a banker says, “I only think this advertising audience is worth a nickel based on past data … but then if an advertiser/company pays 10 cents, the market pays out 5 cents (with a portion going into the Shapley Value Formula as a market fee).</p><p>The important part of this arbitrage auction is the balancing act that a ETF market creates and how the data is public — as the ETF market makes money, advertisers drop their bids automatically to save money. The banker is not betting on a company or any audience, they are betting on the value of a piece of data that a company or individual created, and the potential for that piece of data to match up to an ad’s metadata and pay out via the Shapley Value formula. The optimized matching of the data creates value, not any one piece of the data itself.</p><p>Through this process of bidding, matching, payouts and renewal of the inventory in the marketplace, optimization at a global scale can occur so that advertisers only add a modicum of whitelist/blacklist targeting customizations, and the arbitrage bidding would optimize the rest.</p><p><a href="https://www.bloomberg.com/features/2016-etf-files/toy/"><strong>Bloomberg describes this automated bidding</strong></a><strong> in an ETF as a stabalizing factor for price:</strong></p><blockquote>This is an opportunity for you, the bank — a balancing act. If the ETF’s price is above its value, you can make money by adding more bits of stock to the fund: Buy the stocks for the cheaper price, sell the ETF for the higher price. You can also do the opposite: If the ETF’s price is below its value, you can return shares of the ETF in exchange for the stocks it stands for, which you can sell for more.</blockquote><blockquote>The ability to create and redeem shares gives big investors and market makers an incentive to arbitrage the price of the ETF with the fair value of the stocks in the basket. This continuous arbitrage is why an ETF’s price on the exchange is typically very close to the value of the securities in the basket. <strong>So you get paid to get things more balanced; the difference is your prize money.</strong> This makes the ETF more useful and reliable to everyone. Congratulations, you’re a market-maker arbitrageur!</blockquote><h4>Here’s a visualization of this balancing effect in an ETF Market:</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-yVwdbYawXJWHZG7KPPy4A.gif" /><figcaption>From <a href="https://www.bloomberg.com/features/2016-etf-files/toy/">Bloomberg</a></figcaption></figure><h3>Picking Winners and Losers in the Cookie Exchange</h3><p>Every website is different, and different types of users visit different types of content — but every page view generates value for someone. Every time a webpage loads, data is being sent to companies who monetize it. This is important to understand — small amounts of money are generated from even just one user loading one webpage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0FmgqY7U9-CYTAveGcI0Ow.jpeg" /></figure><p><strong>The important question is: how is that value organized for a market?</strong></p><p>In the Cookie Exchange market, the data that is appended to a user when they load a website is then passed into advertising platforms, and based on the targeting and bidding of the ads auction, a user is shown a specific ad. When the targeting metadata in the ads match to user metadata in a pseudonymized cookie, those publishers providing each piece of data are paid out from a user clicking on that ad based on the total number of matching pieces of metadata from the ad click.</p><p>The combination of billions of websites and thousands of potential targeting criteria creates a complex web of value, but it’s still finite. The question then becomes, what makes a cookie less valuable? How do you determine investment strategies?</p><h4>Indexes and Baskets in a Cookie Exchange</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bz_PSW7zIXrendC9jHQOFQ.jpeg" /></figure><p>In order to make Cookie Exchanges both approachable to the general public but also optimizable by banks and professional investors, it will be essential to have Indices, or groups of stocks (like Dow Jones, S&amp;P 500) and then also have Baskets, or ad-hoc groups of cookies bundled by analysts and audience amalgamators, in order to provide the diversification of value and risk. Many Data Management Platform (DMP)/Demand-Side Platform advertising companies may end up being significant players in the basket creation process.</p><p>Certain audiences and buying behaviors could cross exchanges via cross-ETF-market arbitrage. If Europe creates an exchange and the U.S. and China, there could be cross-exchange arbitrage to stabilize the global cookie prices. A secondary benefit of this cross-exchange arbitrage is the slowing down or reversing of <a href="https://qrius.com/the-splinternet-is-coming-why-countries-will-break-away-from-todays-internet/">the so-called “splinternet</a>.”</p><h3>Cookie Exchanges — Retargeting Refunds, Automated Lookalike Audiences &amp; Targeting, Reduced Click Costs, “Shopping Mode” and Other Reasons to Get Excited about the Future of Advertising</h3><p>I’ve been a digital strategist for over 11 years (first campaign was 15 months on President Obama’s 2007–2008 digital team) and have bought millions of dollars in ads during my career. I currently manage tens of thousands in Facebook and Google ads per month with my team, and spike more during certain seasons and launches.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*Aq0slNNOB0ph14quXhyS4g.gif" /></figure><p>I love both Facebook and Google for a lot of reasons — that’s where the majority of my budget goes (with smaller amounts through Kenshoo for Bing/Yahoo mirroring, and some other one-off networks). Google’s and Facebook’s data is the foundation of online advertising strategy for most businesses and will be for some time… But it doesn’t mean we can’t do better. And it definitely doesn’t mean that we need to remain married to a market essentially comprised of two competitors.</p><p>Cookie Exchanges open up dozens of new advertising strategies and ways to monetize audiences, grow businesses, all while protecting user privacy. Here are just a few of the ways advertisers and existing market players will be able to benefit from Cookie Exchanges…</p><h3>Retargeting Refunds &amp; Ending the Self-Defeating Retargeting Monopolies</h3><p><strong>When a company drops a cookie for a user, why does that data stay siloed? </strong>Companies would be wise to look for an open market that allows automatic expansions of audiences across networks — costs would drop, new leads would be easier to find, and walled gardens would stop artificially increasing prices for highly-targeted ads. This also promotes competition by enhancing access to the data market.</p><h4>Finding more active buyers through pseudo-cookie transparency</h4><p>When a visitor on your website puts an item in their cart and then does not purchase, that visitor is much more valuable on an advertising network than your average website visitor. Marketers called these opportunities, “Abandoned Carts” and they include people who are in the final stages of buying decisions, and thus more valuable to a business.</p><p>Advertising networks are making huge profits off companies that don’t realize that Abandoned Cart Retargeting ads use 100% of that companies data for the ad targeting. Retargeting ads are normally very cost-effective for marketers when compared to other ads, but the data that the company is providing the advertising firm is massively more valuable than the discount they receive on the ads.</p><p>It’s essential for publishers and product owners to take back the data they have about the intent of buyers, and start to monetize their abandon cart visitors, and the exclusion data for people who make purchases (aka they are no longer in the market for X product category).</p><p>In the Cookie Exchange, an Abandoned Cart visitor who did not complete their purchase has a high intent to potentially click on an ad of a competitor (or the original product owner) and complete a purchase. Therefore that grouping of individuals (baskets of cookies) has a higher value on the advertising open market exchange due to their increased likelihood of engaging with an ad that matches their user data. If someone in the ETF Market is betting on the futures of that basket of retargeting cookies to have consistent value stacked against other cookie properties, that would likely be a good investment.</p><h4>My data, my people — the retargeting strategy fallacy</h4><p>Imagine this classic retargeting scenario — currently, when you visit a website for a bicycle company and go through the process to shop for that bicycle and then almost complete a purchase, you will only start seeing ads from that bicycle company — “they own the retargeting data.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/295/1*lrqTknvW2dyIQ_xGLYdBFQ.gif" /><figcaption>The public has made it clear they don’t like being “followed around the internet” with advertisements, yet the core concept of retargeting hasn’t changed.</figcaption></figure><p>The current retargeting strategy results in users being “followed around the internet,” which used to be a novel marketing strategy and has since backfired on the entire advertising market as users revolted to the obnoxious and repetitive experience.</p><p><strong>Imagine again for a minute a different retargeting environment</strong> — if you visited that same bicycle website and added a bike to your cart but then didn’t purchase. In the new Cookie Exchange, your pseudo-cookie would show a high-intent to get a new bike, and every company selling bikes would be able to target ads to you with their best bike deals — a user wouldn’t just see one company, they would see real competition among businesses for their sale. This increase in ad competition would provide users a better experience, and prevent companies from burning dollar after dollar in misguided attempts to re-convert their own audiences.</p><p>This new retargeting market creates a new form of competition at the final stage of an advertising conversion and encourage competitors to compete on the open market. A company could even find a business model to provide comparisons for products, with the knowledge that accurate information reviewed by audiences prior to a purchase could still reward that publisher after any final click-advertisements prior to purchase.</p><h4>Automated Audience Expansions without Facebook/Google’s “Lookalike” Audience Fees</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*p_jEnnvfujpvJ9oOwdP9ng.gif" /><figcaption>How do you automate creating lookalike audiences?</figcaption></figure><p>Advertisers need to expand their audiences, and they need to stop trying to target their campaigns themselves. There are huge benefits to not targeting your advertising campaigns yourself — and to end the retargeting monopoly so that businesses can start to get retargeting ad refunds when they provide the audience data for a retargeting campaign.</p><p><strong>Another example scenario:</strong></p><p>User X wants to buy a BMW — goes on their website and shops around.</p><p>User X gets to the BMW checkout but doesn’t buy a car online — she stops and doesn’t give an email either.</p><p><strong>Does that user belong to BMW?</strong></p><p>Do users get any actual benefits when ads “follow them around the internet?”</p><p>When retargeting first came into the mix, people thought it was magic — and it worked really well to drive huge revenue. Everyone was doing it. But the strategy is saturated and the world-at-large hates the way retargeting ads follow you, and as soon as you mention a brand you start to see those brands ads. That’s due to how pixels and data are shared right now — there isn’t an actual public auction. When you say, “I may want to buy a mid-size luxury vehicle” but only visit the BMW website, you actually aren’t telling Mercedes or Aston Martin or Tesla that you’re shopping for a car. So all those car companies, that may have a car that the consumer <em>loves</em>, and they may have an amazing advertising campaign going on… but that user will <strong><em>never</em></strong> see those ads because they only see the BMW retargeting ads. Their shopping intent wasn’t shared in a pseudonymized way for all car companies to know and try to get their business.</p><p>The reality is that there is no individual or organization who can efficiently optimize targeting manually — it’s all hacks and tricks — and there is always a smaller segment or a missing segment that could be applied to a campaign — always. Show me someone who thinks they can optimize targeting for an ad campaign better than an ETF Market and thousands of professional optimization experts writing targeting rules and algorithms, and I’ll show you someone who doesn’t understand the power of AI and how advertising is changing.</p><h4>The New “Shopping Mode”</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*eGT81wPBmYRycNbnwFilPA.gif" /><figcaption>Ugh, as-if users wouldn’t go for a temporary “Shopping Mode” — instead of the current processes where the ads follow them around the internet indefinitely and provide that overwhelming sense of ickiness.</figcaption></figure><p>Small business and large businesses are hemorrhaging opportunities and missing out on people who actively want to be shown more options. Why can’t a user just add one note to their pseudo-cookie “Shopping for a car between $40k-$60k” and start to see dozens of ads from car companies trying to get their business? Why does a user not have full control over when they want only 1 company to send them ads, and when they want all companies to not know their name, but to try and get their business?</p><p>Audience targeting must eventually be automated, and advertising bidding must be done in a public way to drive down costs for businesses. The closed-door advertising arbitrage at Facebook and Google and all the other major companies must end. Public ETF Cookie Exchanges tied to Shapley Value Markets to reward publishers is the only way to fix the current market imbalance while protecting user privacy.</p><h3>Conclusion — Who will build the first Cookie Exchange?</h3><p>GDPR has created a forced market shift by effectively stripping PII data of its value. In it’s stead, pseudonymized data will emerge as the new format to trade data beyond its original use. The value of that data and the market for it can and should be evaluated through a variation of the Shapley Value formula, whereby different “players” contribute, in-cooperation, unequally, but whereby that cooperation ensures the same or more gains as if the player acted alone.</p><p>Pseudonymized data can be traded through ETF markets of pseudo-cookies, which help to stabilize costs and promote competition of online ads.</p><p>Arbitrage of those ETFs will allow for automated targeting optimization and lookalike audiences while also complying with the privacy requirements of GDPR. This, in-turn, enhances regulatory stability and inhibits advertising fraud. New value, ie. revenue, will be generated by ETF markets and paid out to content creators, audience amalgamators and publishers.</p><h3>Video Summary of the Cookie Proposal</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FNKkXCHBW2d8%3Ffeature%3Doembed&amp;url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DNKkXCHBW2d8&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FNKkXCHBW2d8%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/61ee9aacd458ed09ff44080711b13569/href">https://medium.com/media/61ee9aacd458ed09ff44080711b13569/href</a></iframe><h3>Questions that still need to be answered about Cookie Exchanges:</h3><p>1.) How does Ads.Txt play in the Cookie Exchange? The concept of ads.txt seems outdated and exposes competitive intelligence, but there are some benefits to transparent inventory management.</p><p>2.) How can the publishers considering their own walled gardens and unified login systems (particularly in Germany) reconcile their strategies with a pseudonymized Cookie Exchange?</p><p>3.) What’s the best way for a Cookie Exchange to facilitate <strong>pre-click</strong> bidding for Server-to-Server cached ad delivery (aka faster page load times)?</p><p>4.) The only way for the ETF Market and Shapley Value Formula to cooperatively exist is for the ETF Market to pay fees that are distributed to participants. What fee structures would provide $1 billion in new investments distributed through Shapley over 2 years at current data rates and expected ETF trade rates? What other rate factors should be considered?</p><p>5.) Currently the Shapley Value Formula proposed in the example Cookie Exchange had 65% of the click payout going to the network that hosted the ad (Google, Facebook, etc.) and then 35% being paid out through the Shapley Value formula to all the organizations that participated in creating data that matched the ad targeting. If the Cookie Exchange also includes an advertising/publishing DSP and hosts ads, 100% of the click value could be split with the Shapley Value Formula and the last click would just be rewarded a larger share. <em>It’s my opinion that this current proposed 35% audience sharing rate should be artificially large at the outset of the exchange to provide a runway for audience amalgamators and current market players to continue to monetize their current assets while pivoting into the new market. I believe there could be a benefit to graduated rates that slowly reduce the 35% audience rate until the market would stabilize, and capping audience/publisher/content fees to about 25% of 3rd party ad revenue seems consistent with the impact of that audience data. The breakdown of the SVF and the significance of a “cookie contribution” needs more debate! :)</em></p><p>6.) What types of insider trading rules need to be established? How can automated rules help to find bad actors in the current market?</p><p>7.) Automated javascript pixel deployment (and SDK) error-checks are infinitely easier with a consistent advertising market like the Cookie Exchange. What types of businesses need to be developed to provide regulatory and compliance support for accurately and trusted cookie deployments? A good analogy would be meter-readers and other industries that have processes to always ensure they are accurately recording data.</p><p>8.) How could the Cookie Exchange facilitate a new type of “knowledge-based advertising” that would reward news organizations?</p><p>9.) How could the Cookie Exchange help to reclassify certain types of data as more valuable than they currently are considered?</p><p>10.) Companies like SAP <a href="https://support.ca.com/cadocs/0/h005111e.pdf">who already have pseudo-cookie deployments</a> seem to be ahead of the curve for operational challenges. What other industry organizations need to be contacted for feedback on the Exchange to ensure that work isn’t being duplicated?</p><p>11.) Facebook has over 1,300 user properties used for advertising targeting. How will the Cookie Exchange define quality data, and what caps on the number of properties per cookie session will be established, and how will the user property targeting schema be standardized?</p><p>12.) How could the Shapley Value provide data for a new search engines?</p><p>13.) Purchase exclusion audiences are infinitely more valuable than an inclusion audience and could be dominated by Amazon. How can this audience be properly valued to ensure that more companies have an incentive to improve the quality of their post-purchase data?</p><p>14.) There’s nothing stopping users who share data from also being paid via the Shapley Value formula — this is particularly true for “Shopping Mode” where a user inputs what they want and then gets bombarded with aligned ads trying to get their businesses. All Cookie Exchanges will need to determine if user payouts or dividends are essential to global market participation and more consideration for user-desires should always be a goal.</p><p>15.) What growth decay formulas will be best suited to predict cookie expirations in the ETF market?</p><p><em>Got other feedback or a fun project? Hit me up on Twitter </em><a href="https://twitter.com/thezedwards"><em>@thezedwards </em></a><em>on </em><a href="https://t.co/CPRN5NRWoZ"><em>Linkedin</em></a><em> or via my company website @ </em><a href="https://victorymedium.com/"><em>VictoryMedium.com</em></a><em> if doing cool work✌</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*S01lAwVjrDAy-8cbUGCaVw.gif" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=592cbde9b44b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Would Google’s Artificial Intelligence Director Leave for Apple?]]></title>
            <link>https://medium.com/@thezedwards/why-would-googles-artificial-intelligence-director-leave-for-apple-57329b1d10d2?source=rss-e55d568b3a27------2</link>
            <guid isPermaLink="false">https://medium.com/p/57329b1d10d2</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[gdpr]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Zach Edwards]]></dc:creator>
            <pubDate>Tue, 08 May 2018 02:32:13 GMT</pubDate>
            <atom:updated>2018-07-20T16:13:22.437Z</atom:updated>
            <content:encoded><![CDATA[<h3>Why Would Google’s Artificial Intelligence Director Leave for Apple? The Race for NLP, 3rd-Party Audio Dark Metadata Privacy Concerns, the May 25th Implementation of GDPR, and Apple’s HAL</h3><p>Apple, Google and Amazon are currently locked in an epic quest over the following multi-trillion dollar question: <strong>which</strong> <strong>company will be the first to maintain 99% accuracy for speech recognition from Natural Language Processing (NLP) </strong>in any and every scenario and context, for every language, dialect and slang?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/1*BXcgzdLk7TxeKv7DOLd8gw.gif" /><figcaption>Google AI <strong>vs</strong> Apple AI <strong>vs</strong> Amazon AI</figcaption></figure><p>Beyond just accurate intent and entity matching, there is another race in speech recognition<strong> — the race to organize and categorize audio dark metadata.</strong></p><p>There are rarely discussed consequences of audio dark metadata for the privacy of 3rd parties who haven’t agreed to the Terms of Service of the company providing the audio processing — the un-agreed person standing on the train talking near a person dictating into a Google Assistant-powered app, or the co-worker talking about business as someone nearby asks Siri for directions to an after-work dinner, and many other potential scenarios.</p><p>There is a huge value for a business to acquiring and analyze dark audio metadata, and it’s important for the public and governing bodies to understand their value.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/1*gDldKcrXJf1oMlkfZ7Hgiw.gif" /></figure><p>The winner of the speech recognition race will be the company that not only perfects their segmented NLP models, but also the one that does it under the legal, ethical and social norms required in the marketplace.</p><p>Apple appears to have a dominant position with their rarely-reported-on Hardware Abstraction Layer (“HAL”) architecture which attempts to process and provide some layer of anonymization directly on their devices before the data is stored. In contrast, Google, Amazon Microsoft and other companies appear to currently skip the on-device anonymizing step. Instead, their public statements and developer documents indicate that their audio Natural Language Processing (“NLP”) tools send all recorded audio to their servers, store it, and then don’t take additional steps to remove private 3rd party audio.</p><p>Indeed on Google, Amazon and Microsoft’s websites explaining their speech to text software, they make no mention of third party privacy rights or 3rd party audio dark metadata protection at all. Whether these companies have some work around for filtering and protecting third party metadata once on their server, remains to be seen.</p><p>What is abundantly clear, however, is that on May 25, 2018, the <a href="http://eur-lex.europa.eu/legal-content/en/TXT/?uri=CELEX:32016R0679">EU’s General Data Protection Regulation (GDPR)</a> becomes effective, and creates a whole potential universe of additional liability with regard to the collection of third party audio dark metadata. Though most in the tech world have focused on <a href="https://comm.ncsl.org/productfiles/95595057/FPF_Always_On_WP.pdf">first party data retention policies</a> (terms of use, individual user privacy rights, enforceability of privacy agreements), protecting third party audio dark metadata — that is, data collected and recorded from people and things who are <a href="https://www.scientificamerican.com/article/pogue-how-private-is-your-voice-assistant-device/">not the first party users of the speech device</a> — marks the new frontier in privacy. With new regulations and privacy concerns growing daily, the ultimate winner of the speech recognition race will not only be the one who makes it to 99% recognition the fastest, but the company who successfully does this while also protecting the metadata of first and third party users.</p><p>Time is of the essence, due to GDPR going into effect on May 25th, 2018, and if a company doesn’t comply, the <a href="https://www.eugdpr.org/eugdpr.org.html">EU’s GDPR clearly states</a>, <strong>“those organizations in non-compliance may face heavy fines.” </strong>The formula to determine GDPR fines could cost American businesses potentially hundreds of millions or billions of dollars, and as of publishing this article, there will be less than 3 weeks before the law goes into full effect.</p><p>If 3rd party audio dark metadata rights are included under GDPR, many companies need to make significant changes to their speech recognition NLP systems in a short period of time.</p><h4>The Audio NLP Race for Enterprise Deployment + Audio Dark Metadata Collection and Filtering</h4><p>Apple, Google, Amazon and dozens of other companies are in a race to perfect their Audio Natural Language Processing (NLP) systems and deploy it for enterprise clients and the public at-large.</p><p>Both Google and Amazon are moving forward with nearly identical strategies to create open audio NLP API’s so that 3rd party developers integrate their hardware with Google or Amazon NLP API’s — a strategy to win the race through enterprise deployments, which is the classic enterprise growth strategy. These integrations use the microphones in devices that record anything and everything and pass all of that audio data into Google and Amazon servers — scooping up all 3rd party audio dark metadata without any attempt to filter on-device.</p><p>Apple is approaching audio NLP radically different, with a focus on their on-device audio dark metadata collection and parsing to potentially protect user privacy, through their aptly-named “Hardware Abstraction Layer (HAL)” on all devices with Siri.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*JwLY56fH_YXSc2101pTU_w.gif" /></figure><p>Apple modeled their on-device processing on some similar techniques that <strong>industrial NLP systems with limited-dictionary speaker-dependent processing</strong> have used. This is the type of industrial tech that powers voice activated bar code readers someone could use while working on a dock in Baltimore that has 50–60 core commands that are trained by each person. Apple is working under the same concept but just with one attempting to do it for every word in every language. <strong><em>Easy!</em></strong></p><h3>What is Dark Metadata and Why Does it Matter? The Hidden, Unstructured Data Needing to be Organized Before a Value Can be Defined</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*KL1haDG5npE8ttT1aTVtwQ.gif" /></figure><p>Dark metadata is a machine learning (ML) concept that is slightly hard to understand — basically it’s the concept of a firehose of unique and outwardly indecipherable data and then finding a way to organize that data so that <strong>the output has a new and distinct value</strong>.</p><p>Consider the following analogy: you want to come up with a way to identify every car in need of an oil change. But the only information you have is that you’re standing on a freeway overpass, trying to look at the make and model of every car that whizzes by you. For one person, that data collection and analysis would be impossible, but with the right visual and audio sensors and with enough training data and supplemental user data, the combination of the make, model, time, exhaust, speed, and temperature of every passing vehicle could create a business automation that would allow a business to advertise discount coupons for an oil change at the perfect time. <strong><em>Magic!</em></strong></p><p>So, what does dark metadata have to do with the race for 99% speech recognition? Well, just about everything. It’s essentially the key that will unlock that final piece of the speech recognition puzzle, and Apple, Google and Amazon (and many other companies) know it.</p><p>For Apple’s part, it would at least outwardly appear they are focusing on identifying audio dark metadata and likely are trying to remove 3rd party metadata or anonymize it as background noise, but Google and Amazon have prominent sales and development resources stating that they store their audio Natural Language Processing (NLP) API data on servers prior to any analysis, making it literally impossible to remove 3rd party audio dark metadata prior to storage.</p><p>Google and Amazon (and other similar companies) aren’t hiding that they store all audio data before analysis, but they haven’t explained how they prove they are deleting 3rd party audio dark metadata that has been acquired outside their Terms of Service. All companies involved collecting 3rd party audio dark metadata should provide more details to the public, and everyone should likely figure out their own policies before GDPR goes into effect on May 25th, 2018.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*1hQIoKqGcfnUMfKrw_9pbQ.gif" /></figure><h3>Apple’s 2015 Acquisition of Perceptio Paved the Way for Device-Centric AI with Local Device Processing to Strengthen User Privacy</h3><p>A largely under-the-radar AI acquisition from Apple came in 2015 when they purchased Perceptio, <a href="https://www.macrumors.com/2015/10/05/apple-acquires-ai-startup-perceptio/">as reported by MacRumors who quoted the original Bloomberg piece</a>:</p><blockquote>Perceptio’s goals were to develop techniques to run AI image-classification systems on smartphones, without having to draw from large external repositories of data. That fits Apple’s strategy of trying to minimize its usage of customer data and <strong>do as much processing as possible on the device.</strong></blockquote><p>Over 3 years ago, Apple was already laying the groundwork for a privacy-centric NLP AI, and these steps to double-down on audio dark metadata with on-device processing have opened up untold strategic doors.</p><p><strong>One way to think about the local processing of 3rd party audio dark metadata is through a simple scenario — imagine 20 people are sitting on a train and 12 people have iPhones and 2 of those people are asking Siri a question — 8 people total are talking. </strong>Siri could identify the 2 “primary user” speakers who have agreed to the Apple Terms of Service and start to identify all the background noises as “private 3rd party audio dark metadata” — everything from the known-train-whistle-sound (limited privacy concerns) to the knew and dynamic conversations of other people on the train (significant privacy concerns). In order for the privacy of the six-other people talking who aren’t using Siri and potentially not even Apple customers to be respected and not have their words and conversations documented, Apple’s 3rd-party audio dark metadata processing would be identifying that in real-time, and anonymizing any of the 3rd party audio dark metadata and then passing only the 1st-party audio metadata into Apple servers for user-identified storage and analysis. It’s possible that 3rd party user audio could be completely deleted if algorithms could identify words that would compromise another non-Apple user’s privacy.</p><p>Apple still needs to clarify publicly how they manage 3rd party audio dark metadata — and although they may be architected with Siri + HAL in a way that appears to more strongly support user privacy, they still need to proactively communicate to the public what they’re doing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*k5HBIRx0hBhvpbPsqCBe1g.gif" /></figure><p>If audio streams aren’t being locally processed so the device can identify 3rd-party audio dark metadata before transmission to an off-site server, then that work needs to occur on-server after remote processing. If a company doesn’t filter 3rd party audio on-device, then, currently, the public at large has to trust companies to delete or anonymize/pseudonymize the 3rd-party audio dark metadata once they have it on their servers.</p><p>For both Google and Amazon, they provide free and extremely cheap NLP API services — and as the saying goes — if something is free, you’re the product.</p><p>The benefits for Amazon and Google to provide free and cheap speech to text processing services are greatly over-compensated by their collection of users’ 1st party metadata and 3rd party audio dark metadata. The same goes for Apple and any company working in this space — the public needs these services to develop new technology, but the 3rd party audio dark metadata needs to be acknowledged and protected — and policies to protect this data need to be communicated publicly.</p><h4>Apple’s Spring 2017 Acquisition of Lattice Data, Dark Metadata Experts Who Created Stanford’s Open Source Deep Dive, Vital to Hal</h4><p>As <a href="https://appleinsider.com/articles/17/05/13/apple-acquires-dark-data-specialist-lattice-data-for-200m">reported by Apple Insider after the May 2017 acquisition of Lattice Data</a>, the goal of using dark metadata to train distant supervision machine learning is rooted in saving time and pushing speed and scale of deployment — with the initial privacy implications being somewhat ignored:</p><blockquote>The company contrasts its approach to traditional machine learning, explaining <strong>“we do not require laborious manual annotations. </strong>Rather, we take advantage of domain knowledge and existing structured data to bootstrap learning via distant supervision. We solve data problems with data.”</blockquote><blockquote>The firm also emphasizes the “machine scale” of its platform in being able to “push the envelope on machine learning speed and scale,” resulting in “systems and applications that involve billions of webpages, thousands of machines, and terabytes of data” at what it describes as “human-level quality.”</blockquote><p>As Apple has planned their HAL/SIRI speech to text privacy architecture for years, both Amazon and Google have been focusing on open speech to text networks and growing their data ingestion through enterprise deployments and the public-at-large. <em>Only time will tell which strategy was correct.</em></p><h3>Amazon’s Audio NLP Market Leadership — 3rd Party Audio Dark Metadata Stored via “Amazon Comprehend Voice of Customer Analytics”</h3><p>Amazon built what they call “<a href="https://aws.amazon.com/comprehend/">Amazon Comprehend</a> — Voice of Customer Analytics” — their Natural Language Processing (NLP) software available for developers working with audio to text.</p><p><strong>Here’s the chart that Amazon </strong><a href="https://aws.amazon.com/comprehend/"><strong>has on their developers portal</strong></a><strong> showing how audio data is stored <em>before</em> the data is parsed or processed.</strong> This proves that any app developers 3rd party audio dark metadata from non-agreed-TOS-parties would be transferred and stored by Amazon. The only thing not clarified here is if Amazon removes or anonymizes any data in the processing stage or at any time period afterwards.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VjJ5AmU-i_aCTLWX8PUdKQ.jpeg" /><figcaption>Amazon Comprehend Image from <a href="https://aws.amazon.com/comprehend/?sc_channel=PS&amp;sc_campaign=aiml_pmm_q22018&amp;sc_publisher=google&amp;sc_medium=ai_machine_learning_q22018&amp;sc_content=nlp&amp;sc_detail=%2Bnlp&amp;sc_category=&amp;sc_segment=267777410550&amp;sc_matchtype=b&amp;sc_country=US&amp;ef_id=WutWiwAACn22UPyQ:20180504215508:s">AWS.Amazon.com</a> — pink text and arrows added for context</figcaption></figure><h3>Google has Extremely Strong Speech to Text NLP Developer Tools, No Mention of 3rd Party Privacy from Audio Dark Metadata After Storage</h3><p>Google is a dominant force in AI, particularly with <a href="https://cloud.google.com/speech-to-text/">their Google Cloud Speech to Text developer tools</a>. They have a global babel-fish API that can ingest audio in 120 languages — and has a variety of other labeling features:</p><blockquote>Google Cloud Speech-to-Text enables developers to convert audio to text by applying powerful neural network models in an easy to use API. The API recognizes 120 languages and variants, to support your global user base. You can enable voice command-and-control, transcribe audio from call centers, and more. It can process real-time streaming or pre-recorded audio, using Google’s machine learning technology.</blockquote><p>Google has never shy’d away from asking for raw audio via file uploads, API connections, web microphones, and via basically any hardware/software that can pay the Speech to Text fee of $0.006 USD per 15 seconds of transcribed audio — and the first 60 minutes is always free. Google has also long been a leader for user-centric privacy controls and lets users listen and delete all their recordings sent to their services. They haven’t however, in any significant way, explained their 3rd party audio dark metadata retention policies and any changes in their policies due to GDPR.</p><p><strong>Here’s Google’s public Cloud Speech to Text webpage showcasing the features that ensure 3rd party audio dark metadata is always sent to Google prior to any sort of on-device parsing:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jRGmlVCimqqmYI8JNZne8Q.jpeg" /><figcaption><em>Google cloud Speech-to-Text Image from </em><a href="https://cloud.google.com/speech-to-text/"><em>Cloud.Google.com</em></a><em> — pink box and arrows added for context</em></figcaption></figure><h4>Microsoft’s Bing Speech API Similar to Google and Amazon — Firehose Audio Ingestion of 1st and 3rd Party Audio Dark Metadata</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ny4Mn3hT7VffVxZ_dNNnlA.jpeg" /><figcaption>Bing speech API screen shot from <a href="https://azure.microsoft.com/en-us/services/cognitive-services/speech/">Azure.Microsoft.com</a></figcaption></figure><p><a href="https://azure.microsoft.com/en-us/services/cognitive-services/speech/">Microsoft’s Bing Speech API </a>is structured almost identically to both Amazon and Google. Audio data is sent to Microsoft servers prior to analysis. Their Cortana devices<a href="https://www.wired.com/2016/12/alexa-and-google-record-your-voice/"> were also built</a> similar to Google and Amazon, with the voice data being transferred to Microsoft servers for processing.</p><p>News reports have also pointed out that <a href="http://gazette.com/why-google-microsoft-and-amazon-love-the-sound-of-your-voice/article/1594412">Microsoft’s Xbox has a app they produced called Voice Studio</a> that listens to background sounds that come through microphones as people are gaming on Xbox. The same article also explained several other audio dark metadata uses-cases have been developed by Microsoft and Google, with the Microsoft audio dark metadata collection and filtering in airports having potential privacy implications for business travelers:</p><blockquote>Companies are also designing voice recognition systems for specific situations. <strong>Microsoft has been testing technology that can answer travelers’ queries without being distracted by the constant barrage of flight announcements at airports. </strong>The company’s technology is also being used in an automated ordering system for McDonald’s drive-thrus. Trained to ignore scratchy audio, screaming kids and “ums,” it can spit out a complicated order, getting even the condiments right. Amazon is conducting tests in automobiles, challenging Alexa to work well with road noise and open windows.</blockquote><p><strong><em>NOTE: </em></strong>For the same reason this research doesn’t include every company working with audio NLP (Baidu and Samsung are doing very interesting things), Microsoft was largely ignored due to their current U.S. market share. It is clear they operate in very similar ways to both Amazon and Google, and most news accounts would point to Apple as the outlier.</p><h3>Apple’s Privacy-Centric Strategy for Local Audio Dark Metadata Processing and Filtering via HAL</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KEZDya2vErATNh6_doXCXA.png" /></figure><p><a href="https://www.nytimes.com/2018/04/03/business/apple-hires-googles-ai-chief.html">As reported by the New York Times</a> last month, Apple is focused on Natural Language Processing (NLP) without compromising privacy — something that competitors have balked at:</p><blockquote>Apple has taken a strong stance on protecting the privacy of people who use its devices and online services, which could put it at a disadvantage when building services using neural networks.</blockquote><blockquote>Researchers train these systems by pooling enormous amounts of digital data, sometimes from customer services. Apple, however, has said it is developing methods that would allow it to train these algorithms without compromising privacy.</blockquote><p>Apple provides 3rd party audio NLP developer tools in a very unique way — radically different than Amazon, Google, Microsoft and most of the industry. Apple only provides audio NLP through their hardware and describes it on <a href="https://developer.apple.com/documentation/coreml">their CoreML about page</a>, “<strong><em>Running strictly on the device ensures the privacy of user data.”</em></strong></p><p>Here is a screen shot from the Apple CoreML page describing how they protect user privacy by only providing audio Natural Language Processing through their hardware devices:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gboz37k0f5-nSx7eUYDc5g.jpeg" /><figcaption><em>Apple Core ML screen shot from </em><a href="https://developer.apple.com/documentation/coreml"><em>Developer.Apple.com </em></a><em>—pink box added for context</em></figcaption></figure><p>In the <a href="https://developer.apple.com/library/content/documentation/MusicAudio/Conceptual/CoreAudioOverview/WhatisCoreAudio/WhatisCoreAudio.html#//apple_ref/doc/uid/TP40003577-CH3-SW1">developer resources for Core Audio</a>, Apple names this privacy layer the Hardware Abstraction Layer, or HAL, writing:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*sSE_KV6em7ewz-FbY9E_hg.gif" /><figcaption>“<em>I’m sorry, Dave. I’m afraid I can’t do that.”</em></figcaption></figure><h3>“In OS X, the majority of Core Audio services are layered on top of the <strong>Hardware Abstraction Layer (HAL)</strong> as shown in Figure 1–1. Audio signals pass to and from hardware through the HAL.”</h3><h4>Apple’s HAL — the real product of Apple AI</h4><p><a href="https://techcrunch.com/2017/05/13/apple-acquires-ai-company-lattice-data-a-specialist-in-unstructured-dark-data/"><em>Apple acquired dark metadata company Lattice Data in Spring 2017</em></a><em> and on-device AI processing experts Perceptio in 2015 have been deploying a version of </em><a href="http://deepdive.stanford.edu/"><em>Stanford’s Deep Dive project</em></a><em> in </em><strong><em>their ironically-named Hardware Abstraction Layer (HAL)</em></strong><em>. Lattice and Perceptio IP are core elements of HAL along with likely hundreds of other team-members and IP from other less-visible acquisitions. The production of AI-centric co-processors like HAL has increased in recent years and is likely to continue growth as the full privacy implications of GDPR is seen-out.</em></p><p>Beyond the core privacy benefits, Apple’s HAL architecture provides a platform to create fully segmented speech NLP algorithms that are able to break the speaker-independence curse that occurs from monolithic speech to text training — aka the “not everyone sounds like a developer from San Francisco” conundrum. It’s obviously slowed down their entire development process, but conceptually, their goal is clear.</p><p>This concept was <a href="https://techcrunch.com/2015/09/11/apple-addresses-privacy-questions-about-hey-siri-and-live-photo-features/">explained by Apple to Tech Crunch in 2015</a> when other privacy concerns were raised, with HAL providing pseudo-security for false wake-words:</p><blockquote>Instead, audio from the microphone is continuously compared against the model, or pattern, of your personal way of saying ‘Hey Siri’ that you recorded during setup of the feature. Hey Siri requires a match to both the ‘general’ Hey Siri model (how your iPhone thinks the words sound) and the ‘personalized’ model of how <em>you</em> say it. <strong>This is to prevent other people’s voices from triggering your phone’s Hey Siri feature by accident.</strong></blockquote><p>Apple also explained their use of random IDs instead of personally-identifiable information when sending their Siri data to their servers, which likely is similar to how they handle 3rd party audio dark metadata:</p><blockquote>Of course, as has always been the case with Siri, once a match is made and a Siri command is sent off to Apple, <strong>it’s associated with your device using a random identifier, not your Apple ID or another personalized piece of info. </strong>That information is then ‘approved’ for use in improving the service, because you’ve made an explicit choice to ask Apple’s remote servers to answer a query.</blockquote><p>Here is a screen shot from <a href="https://developer.apple.com/library/content/documentation/MusicAudio/Conceptual/CoreAudioOverview/WhatisCoreAudio/WhatisCoreAudio.html#//apple_ref/doc/uid/TP40003577-CH3-SW1">Apple’s Developer portal</a> showing this privacy and segmentation architecture of HAL within the scope of Core Audio:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BoaPeFpmPiWw6vFNJ8h7Iw.jpeg" /><figcaption><em>Apple Core Audio Screen Shot from </em><a href="https://developer.apple.com/library/content/documentation/MusicAudio/Conceptual/CoreAudioOverview/WhatisCoreAudio/WhatisCoreAudio.html#//apple_ref/doc/uid/TP40003577-CH3-SW1"><em>Developer.Apple.com</em></a><em>— HAL, pink text and arrows added for context</em></figcaption></figure><h4>Apple and 3rd Party Audio Dark Metadata — Hardware Dominance Supports a Privacy-Centric NLP AI Strategy</h4><p>For Apple, post-Lattice and Perceptio acquisitions, the obvious use-case for optimizing massive amounts of audio dark metadata comes from their hardware dominance . iPhone is growing exponentially faster than Android — <a href="https://investorplace.com/2018/04/apple-inc-s-aapl-iphone-x-twice-profitable-android-manufacturers-combined/">in the fourth quarter of 2017, Apple’s iPhone X generated over 5x the profit of 600 Android devices <strong>combined</strong></a>. Also, a recent 2018 poll of U.S. teenagers showed growing demand for iPhone’s with <a href="https://www.barrons.com/articles/apple-teens-say-they-want-iphones-and-watches-1523375050">84% of teens stating their next phone would be an iPhone</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tzLlM6Zu3PPC06WvyibQ9A.jpeg" /><figcaption><em>Screen Shot from </em><a href="https://www.apple.com/ios/carplay/"><em>Apple.com</em></a><em> of 3rd Party CarPlay App Integrations</em></figcaption></figure><p>Further to this point, <a href="https://www.apple.com/ios/carplay/">Apple CarPlay is now available in over 300 car models</a> and 6 aftermarket audio systems — and Apple has been racking up numerous CarPlay partnerships with huge audio app developers including WhatsApp, Spotify, Pandora, NPR and MLB at Bat.</p><p>Google has lagged behind with Android Auto, which still only <a href="https://www.extremetech.com/mobile/267568-google-rolls-out-wireless-android-auto-but-almost-no-one-can-use-it">has wireless available</a> for people using Google’s own phones (wired connections for most other models). Google does however regularly update their <a href="https://www.android.com/auto/">Android Auto website</a> and are obviously committed to growing that sector in a significant way — but through a totally different deployment and privacy architecture.</p><p>With Americans <a href="http://www.businessinsider.com/americans-driving-300-hours-a-year-2016-9">spending, on average over 290 hours per year on average in their cars</a>, it remains a significant opportunity for NLP systems.</p><h3>Apple Files Patent for Audio Dark Metadata Retargeting Ads and Bridges Air-Gap for Offline-Online Retargeting Ads</h3><p>Read all of Apple’s pending patent here to get the full-scope of their real-world to online retargeting technology:<strong> </strong><a href="https://patents.google.com/patent/US20180083724A1/en"><strong>“Accessing radio content from a non-radio source”</strong></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XrqwlIr8h2tQT_ndFBDk2g.png" /><figcaption>This is the audio-equivalent of breaking the air-gap and retargeting online based on sounds in the world. The future is listening…</figcaption></figure><p>If the early 2000’s were dominated by Search and Product Retargeting, the 2020’s are set to be dominated by real-world retargeting — and Apple is starting with radio retargeting.</p><p>They aren’t just deploying “you may have listened to” retargeting, these ads will know the second you stopped listening from one device to encourage picking back up at that exact second on another device — and many other use-cases could be on the horizon…</p><p>Apple created the diagram here showing how they are planning to monetize their iPhone and CarPlay audio dark metadata to encourage someone to “finish listening to x show” when they get out of their car.</p><p><strong>Here’s a flow chart for the logic behind the </strong><a href="https://patents.google.com/patent/US20180083724A1/en"><strong>“finish your show” retargeting feature</strong></a><strong>, as described by Apple:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bpw2VGzCkbq91q6uEyjYTw.png" /><figcaption>Second by Second audio dark metadata drives the IP and this logic</figcaption></figure><blockquote>In response to detecting that a user has not finished listening to a broadcast segment (e.g., a radio segment), an electronic device can automatically identify and access an alternate, non-broadcast source for the same broadcast segment (e.g., a corresponding podcast episode). Using the electronic device, a user can play back the segment from the non-broadcast source, starting playback at the last position of the broadcast segment when the user stopped listening to the broadcast.”</blockquote><p>Apple described this Radio-Retargeting (Where you left off) patent and vaguely described the potential use-cases for both music and talk-radio:</p><blockquote>For example, a media broadcast can include a succession of different media items (e.g., different songs). As another example, a media broadcast can include a succession of segments (e.g., a succession of conversations or interviews with different guests).</blockquote><p>The patent specifically outlines a use-case for Apple’s CarPlay to let a user continue listening to a podcast after they turn off their car and get to their computer:</p><blockquote>A user, however, may not always have access to the radio broadcast stream or media broadcast stream provided by a content provider. For example, communications circuitry or radio tuning circuitry of an electronic device may not receive a radio broadcast (e.g., as the device goes into a tunnel or in an area that radio waves cannot access). As another example, communications circuitry can move too far from a broadcast source (e.g., a communications node of a communications network) to receive broadcast data. As still another example, the user is required to turn off the electronic device (<strong>e.g., turn off a car radio upon leaving a car</strong>). If a user is listening to a particular media item or radio segment of interest, the user may not be able to finish listening to the segment, and can miss parts of broadcast content before a second electronic device can receive the broadcast stream (<strong>e.g.,</strong> <strong>the user turns on a radio in their house or office</strong>).</blockquote><h4>Previous News Reports Confirmed the 1st-Party Processing and Privacy Architecture of Apple, Amazon and Google — Less Clear Policies on 3rd Party Audio Dark Metadata or “Background Noise”</h4><p><a href="https://www.wired.com/2016/12/alexa-and-google-record-your-voice/">Wired’s December 2016 article</a> on speech NLP AI reported that Apple uses a random string of numbers with 6-month data retention, whereas Google and Amazon were storing all audio data until a user decides to delete them. The original article did not mention 3rd party data so it’s unclear what happened to the dark metadata in both of these scenarios. It’s clear however that Apple has taken a divergent approach to audio AI privacy:</p><blockquote>While <a href="https://www.wired.com/2013/04/siri-two-years/">Apple logs and stores Siri queries</a>, they’re tied to a random string of numbers for each user instead of an Apple ID or email address. Apple deletes the association between those queries and those numerical codes after six months. Your Amazon and Google histories, on the other hand, stay there until you decide to delete them.</blockquote><h3>Audio AI Breaking the Air-Gap — 3rd Party Audio Dark Metadata is Worth Billions</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FxThtatVgZVprKd3UEU%2Ftwitter%2Fiframe&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fglas-2018-xThtatVgZVprKd3UEU&amp;image=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2FxThtatVgZVprKd3UEU%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="435" frameborder="0" scrolling="no"><a href="https://medium.com/media/004460f7aa540bd1f35b802751a5cb74/href">https://medium.com/media/004460f7aa540bd1f35b802751a5cb74/href</a></iframe><p><strong>Here’s a scenario — for all intents and purposes, let’s assume that Amazon doesn’t convert background audio into dark metadata on-device and reject any non-primary-speaker audio prior to that data going into their servers (or perform some type of anonymization after being stored).</strong></p><p>Picture for a moment, someone is sitting at their office desk listening to some music from their Amazon Echo speaker they purchased using their business email address — they decide to ask Alexa to switch to Crazy in Love from Beyoncé — and at the same moment they are asking to load up one of the great jams of our generation — someone in their office two cubicles down yells out, <strong>“Ya but it’s HQ’s fault we aren’t going to make the release this quarter!” </strong>If that audio snippet were passed into Amazon’s servers, with the geotagging of the speaker putting it at the businesses satellite office, any audio dark metadata AI models and translations would have all the information it needed to determine that a public company was about to miss revenue projections.</p><p><strong>If machines can hear 3rd party audio dark metadata that includes corporate secrets, and if machines are also buying stock from public data, what protections will keep the machines from trading stocks on the 3rd party audio dark metadata?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*G1wO5D1hcVf632GuOKjzUw.gif" /><figcaption>Audio ML systems will require dark metadata privacy filters aka “Build audio AI like the dude would build audio AI.”</figcaption></figure><p>Another potential scenario is that in any large crowd, if someone activates Siri or any other speech AI system that records background noise, there are people being recorded who have not agreed to the Terms of Service of that company — and if that data is then being parsed and passed directly back into the servers of that company without any effort to protect the speech of a non-user — that’s the type of unintended surveillance that will eventually have some sort of “Cambridge Analytica” moment that forces Congress to step in and regulate the storage and processing of 3rd party audio dark metadata. This moment of regulation is the moment that a privacy-last NLP ML/AI system will be forced to go completely offline and rebuild the business and training data model from scratch.</p><p>There is a significant moat around speech recognition and the challenges to build speech personalization at scale while protecting privacy, and as the race gets further along, we’ll start to see these moments where AI and privacy collide, and the public will see which companies had the foresight to prepare for those moments.</p><h3>April 2018 — John Giannandrea Jumps Ship from Google AI to Lead Apple AI</h3><p>Last month, Google’s AI chief <a href="https://www.nytimes.com/2018/04/03/business/apple-hires-googles-ai-chief.html">John Giannandrea was unexpectedly hired by Apple</a> to lead their AI work, reporting directly to CEO Tim Cook. From some perspectives, the decision is inexplicable — who would bet on Apple over Google or Amazon in the race for AI dominance? The timing with the global discussion on privacy, user data and implementation of GDPR should make a lot of investors ears perk up…</p><p>In an email to staff members obtained by <a href="https://www.nytimes.com/2018/04/03/business/apple-hires-googles-ai-chief.html">The New York Times</a>, Mr. Cook reported on the hire stating, “Our technology must be infused with the values we all hold dear…<strong>John shares our commitment to <em>privacy</em></strong> and our thoughtful approach <strong>as we make computers even smarter and more personal</strong>.”</p><h3>The Wildcard: Inevitable Governmental Regulations, and the May 25th, 2018 GDPR Implementation</h3><p>The <a href="https://www.eugdpr.org/">EU General Data Protection Regulation (GDPR)</a> creates new rights for individuals in Europe and has extremely strict regulations to protect user privacy. It is forcing companies to comply with strict data retention policies and gives individuals new rights of appeal and investigation. When exploring the implications of GDPR, most use-cases have focused on 1st-party data retention, but these policies could apply to 3rd party audio dark metadata and could open up a litany of liability for certain companies providing these very “generous” speech to text NLP API’s for developers while scooping up the audio dark metadata without any attempt to filter out 3rd party data.</p><p>A few of the <a href="https://www.computerweekly.com/feature/GDPR-brings-serious-implications-for-data-storage">new individual rights from GDPR was organized by ComputerWeekly</a> as they analyzed it in the run-up to implementation last year:</p><blockquote>Under GDPR Article 22, the individual will have the right to challenge the way these algorithms work and the decisions they make. This introduces a new approach to seeking consent, as businesses must get permission to use personal data and to process it in certain ways.</blockquote><blockquote>Article 17 of GDPR introduces the “<a href="https://searchcontentmanagement.techtarget.com/definition/The-right-to-be-forgotten">right to be forgotten</a>”. Unless a business or organisation can show a legitimate reason to retain an individual’s data, that person can request the information is deleted by the business without “undue delay”, which may represent a significant challenge for many organisations.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*-PwItG2TQOAcS0rGB9KvtQ.gif" /><figcaption>Rep. Duncan Hunter vaping in Congress to prove a point about vaping being annoying and just another reason to have low expectations for Congress.</figcaption></figure><p>As the full implications of the GDPR implementation are observed, and as the U.S. continues to debate new privacy and data retention legislation, there needs to be additional discussions about whether companies can be held liable for their 3rd party audio dark metadata collection and retention policies. <em>Should a company be required to remove 3rd party audio data before storage? Should a company be required to remove 3rd party audio data after storage? Where is the line?</em></p><p>It is extremely challenging to acquire, store and analyze trillions of pieces of audio dark metadata in real-time while protecting user privacy. None of the companies mentioned in this article can be completely faulted for ignoring this concept in their public messaging, because technically-speaking it’s still early-days with this type of dark audio metadata classification. Fortunately, on-device, speaker-dependent processing through architecture like HAL is challenging the industry and their “store-first” mentality. Streams of 3rd party audio dark metadata are being collected right now by potentially dozens of companies, with likely very few of them even considering Apple’s on-device privacy-centric approach.</p><p>Congress only recently debated the Cambridge Analytica scandal and data acquisition through Facebook’s app permissions, which has gone on for nearly a decade. It’s unclear if and when they will fully understand audio AI systems and the importance of a company providing clarity about how they handle that type of 3rd party data. Once members of Congress and the EU realize that 3rd party audio dark metadata conversations are currently being captured and stored by other user’s devices without 3rd party consent, someone will need to test this current 3rd party audio dark metadata environment through the agreed-upon legal channels in GDPR and other governmental authorities.</p><h3>What other problems should Apple, Google, Amazon, Microsoft and other speech recognition AI companies work on? A few questions to consider…</h3><p>What dark metadata amalgamation strategies could create unique products? This is the trillion dollar question that will need to be answered for decades.</p><p>Work with co-processors has gone on for years and production since at least 2008 — which microchip manufacturers will be the first to develop and market HAL competitors specifically for it’s privacy features?</p><p>How will Apple ensure that HAL and their privacy-centric device processing strategies stay ahead of the game? Which other device and chip manufacturers are best posed to compete or already architecturally prepared?</p><p>What types of other on-device privacy filtering needs to occur from companies like Apple to protect 3rd party audio dark metadata from being exploited?</p><p>How are companies and courts going to define their own efforts to meet the <a href="https://iapp.org/news/a/top-10-operational-impacts-of-the-gdpr-part-8-pseudonymization/">GDPR’s attempt to create an environment of “pseudonymization</a>”?</p><p>How do 1-party and 2-party recording consent laws (patch-work quilt regulations across the U.S. and world) impact 3rd party dark audio metadata collection and storage? <a href="https://comm.ncsl.org/productfiles/95595057/FPF_Always_On_WP.pdf">This is a great report from Stacey Gray from April 2016 for the Future of Privacy Forum</a> that covered this in some new ways, particularly the impact on data retention due to HIPAA. If 3rd party audio dark metadata could be geotagged to a hospital, does that mean it automatically needs to be rejected?</p><p>Should Congress hold hearings on audio dark metadata and the impact on 3rd-party privacy from NLP AI systems?</p><p>Should companies keep audio of their users when captured on another users device they control (ie family plans?)? Where is the line for 2nd party audio tracking?</p><p>Which company will be the next to follow Facebook, <a href="https://newsroom.fb.com/news/2018/04/data-abuse-bounty/">who announced their Data Abuse Bounty Program</a> on April 10, 2018? How do you find and expose your data abuse and internal data problems?</p><p><em>FINAL WRITERS NOTE: During this reporting, the perceived male gender of “HAL” and perceived female gender of “SIRI” annoyed me greatly. Personally on my Apple device I have a British male Siri voice so that when it does poor translations I’m not yelling at a woman. It’s important for all of these companies working with voice to text AI to recognize the gender bias deeply rooted in their technology and look for ways to prevent their bias — this is crucial to prevent the looming SkyNet future. Everyone working in AI should read about algorithmic bias and algorithmic oppression — one of the most eye-opening and important books on the topic is by Safiya Umoja Noble, </em><strong><em>Algorithms of Oppression</em></strong><em> — please consider picking up a copy on </em><a href="https://nyupress.org/books/9781479837243/"><em>NYUPress</em></a><em>, </em><a href="https://www.barnesandnoble.com/w/algorithms-of-oppression-safiya-umoja-noble/1126190426#/"><em>B&amp;N</em></a><em>, or </em><a href="https://www.amazon.com/Algorithms-Oppression-Search-Engines-Reinforce/dp/1479837245"><em>Amazon</em></a><em>. Dr. Noble </em><a href="https://safiyaunoble.com/"><em>is very active and available to give presentations</em></a><em> — every company working with AI needs to ask for her help.</em></p><p><em>Got other feedback or a fun project? Hit me up on Twitter </em><a href="https://twitter.com/thezedwards"><em>@thezedwards </em></a><em>on </em><a href="https://t.co/CPRN5NRWoZ"><em>Linkedin</em></a><em> or via my company website @ </em><a href="https://victorymedium.com/"><em>VictoryMedium.com</em></a><em> if doing cool work✌</em></p><p><em>(Amazon, Apple, Google, and Microsoft were sent a request for comment on this article over a day before publication. No official comment was received as of publishing but I’ll update this if/when statements arrive. No payment was solicited or made in connection with this article. If any company mentioned has any corrections or problems don’t hesitate to reach out via zach[at]victorymedium[dottttt]com and i’ll get on it asap. Any additional comments that you’d like to have added send to the same email or message me here on Medium. Thanks for reading and thinking about 3rd-party audio dark metadata privacy and product implications.)</em></p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FBwUQ1YtUpzw%3Ffeature%3Doembed&amp;url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DBwUQ1YtUpzw&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FBwUQ1YtUpzw%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/7a9d0c3b10a386b11af27f44f6d6bf70/href">https://medium.com/media/7a9d0c3b10a386b11af27f44f6d6bf70/href</a></iframe><h4><strong>Image credits all link to giphy.com unless noted</strong></h4><p><a href="https://giphy.com/gifs/kochstrasse-cute-illustration-WvuMTf88aV3FQ4odG5"><em>Illustration Blink GIF</em></a><em> by Kochstrasse</em></p><p><a href="https://giphy.com/gifs/the-big-lebowski-jeff-bridges-leon-russom-R7yQtEGaWhW36"><em>Ignore the Big Lebowski GIF</em></a><em> by unknown</em></p><p><a href="https://giphy.com/gifs/glas-2018-xThtatVgZVprKd3UEU"><em>Sad Black Friday GIF</em></a><em> by JON</em></p><p><a href="https://giphy.com/gifs/amber-heard-hunt-eCiVjQiNsTZni"><em>Congressman GIF</em></a><em> from The Count</em></p><p><a href="https://giphy.com/gifs/hal-open-the-pod-bay-doors-please-l41YsxKKVYnucStag">2001 A Space Odyssey GIF</a> from Top 100 Movie Quotes of All Time</p><p><a href="https://giphy.com/gifs/editingandlayout-okay-i-see-disagree-5xtDarwlnNgxVN2oO0U"><em>John Candy from Planes, Trains and Automobiles GIF </em></a><em>by unknown</em></p><p><a href="https://giphy.com/gifs/filmeditor-mean-girls-movie-3otPoyXM2G9VRgROnK?utm_source=media-link&amp;utm_medium=landing&amp;utm_campaign=Media%20Links&amp;utm_term="><em>Regina George Privacy from Mean Girls GIF</em></a><em> by unknown</em></p><p><a href="https://giphy.com/gifs/transparent-dump-part-4rCTVitgzyiKQ"><em>Highway time-lapse GIF</em></a><em> from Funny Junk</em></p><p><a href="https://en.wikipedia.org/wiki/Apple_Park#/media/File:Apple_Park_satellite_view_May_2017.jpg"><em>Apple Park Satellite View image</em></a><em> from Wikipedia.com, image from Planet.com</em></p><p><a href="https://giphy.com/gifs/the-chosen-one-cinemagraph-crane-3oEhmF9zaTI69IIW9G">Crane Shipping GIF</a> by The Chosen One</p><p>Final credit goes to this helpful article from the Siri Team @ Apple, “Hey Siri: An On-device DNN-powered Voice Trigger for Apple’s Personal Assistant” via <a href="https://machinelearning.apple.com/2017/10/01/hey-siri.html">Apple.com</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=57329b1d10d2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>