<?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 Sufyan Muhammad on Medium]]></title>
        <description><![CDATA[Stories by Sufyan Muhammad on Medium]]></description>
        <link>https://medium.com/@sufyankr41?source=rss-48acd99e7c72------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*LggisDhMCOd1tEpCcwDO9w.jpeg</url>
            <title>Stories by Sufyan Muhammad on Medium</title>
            <link>https://medium.com/@sufyankr41?source=rss-48acd99e7c72------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 12:09:53 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sufyankr41/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[9.9% of Apps Asked. The Rest Just Tracked You Anyway]]></title>
            <link>https://medium.com/@sufyankr41/9-9-of-apps-asked-the-rest-just-tracked-you-anyway-a9902ebce310?source=rss-48acd99e7c72------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9902ebce310</guid>
            <category><![CDATA[consent]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[android]]></category>
            <category><![CDATA[apps]]></category>
            <category><![CDATA[tracking]]></category>
            <dc:creator><![CDATA[Sufyan Muhammad]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 19:34:34 GMT</pubDate>
            <atom:updated>2026-11-05T19:34:34.384Z</atom:updated>
            <content:encoded><![CDATA[<p>Every app you’ve ever installed almost certainly talks to a tracker before you’ve tapped anything. This <strong>2021 Oxford/KU Leuven</strong> paper decided to check something simple: when apps legally need your consent before tracking you, do they actually ask?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/902/1*zpEeDBGiKRv-Kk8vGKP4RA.png" /><figcaption>Figure 1: Many trackers provide information on what developers need to know to implement consent</figcaption></figure><ul><li>They pulled 1,297 random free apps from the UK Google Play Store, installed each one on a real phone, ran it for 15 seconds without touching anything, and logged every network connection it made</li><li>71.3% of apps contacted a known tracker before the user did anything at all</li><li>Only <strong>9.9%</strong> of apps asked for consent in any form</li><li>Of the apps that did ask, 43.7% only gave you one option: agree, or don’t use the app. Just 3.5% of all apps gave a genuine “no” that didn’t boot you out</li><li>They also checked the 13 most common tracker companies (Google AdMob, Facebook, Unity Ads, and so on) to see if the tracker makers themselves nudge developers toward asking for consent</li><li>Only<strong> 3</strong> of 13 trackers implement a consent flow automatically, out of the box</li><li>Most tracker companies’ own setup guides don’t even mention consent is needed</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/913/1*Hy3xOZDFyYkyws7fwBA6ww.png" /><figcaption>Figure 2: While most apps on the Google Play Store use third-party tracking, only few apps allow users to refuse consent</figcaption></figure><p>What stuck with me is how little of this comes down to developers being careless. The paper’s framing made that clear: most app developers aren’t privacy lawyers, they’re following a setup guide from Google or Facebook, and if that guide never mentions consent, there’s no reason they’d think to add it. The responsibility quietly shifts onto companies that have every financial incentive not to make opting out easy, and the paper backs this up directly <strong>6</strong> of the 13 trackers studied likely require consent under EU law but neither implement it by default nor mention it anywhere in their own setup instructions.</p><p>The surprising thing was, finding about what happens after someone opts out. Among the small number of apps that offered a real opt-out, tracker contact still barely dropped. People were clicking <strong>“no”</strong> and getting followed almost the same amount anyway, because “no” usually only meant “no personalized ads,” not “no tracking.” That’s a more cynical gap than I expected. It’s not that opt-outs don’t work well, it’s that most of them were never built to stop anything meaningful in the first place.</p><p>A few more details worth pulling out:</p><ul><li>Google was the single most contacted company, reaching <strong>58.6%</strong> of all tested apps, more than Facebook, Unity, and every other tracker combined</li><li>Apps running display ads contacted roughly three times more trackers on average than apps without ads</li><li>Even the trackers’ own compliance guidance was hard to use. Three of the thirteen companies linked to pages that were dead, outdated, or simply 404 at the time of testing</li><li>The guides scored as genuinely difficult to read by standard readability metrics, even before getting into the legal language</li></ul><p>The sort of query that I have is that this paper measured only the first <strong>15</strong> seconds after an app opens, before anyone’s clicked anything. I’d want to know what tracking looks like after someone’s actually used an app for a week, logged in, granted a few permissions along the way. The first 15 seconds is already bad. I have a feeling the honest number, over real usage, is worse, not better. That’s the natural next study to go looking for.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9902ebce310" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[People Check Their Privacy Settings Once, Then Forget]]></title>
            <link>https://medium.com/@sufyankr41/people-check-their-privacy-settings-once-then-forget-a61d1c46a354?source=rss-48acd99e7c72------2</link>
            <guid isPermaLink="false">https://medium.com/p/a61d1c46a354</guid>
            <category><![CDATA[location]]></category>
            <category><![CDATA[data-privacy]]></category>
            <category><![CDATA[mobile-apps]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[push-notification]]></category>
            <dc:creator><![CDATA[Sufyan Muhammad]]></dc:creator>
            <pubDate>Wed, 07 Oct 2026 13:51:09 GMT</pubDate>
            <atom:updated>2026-10-07T13:51:09.328Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wBajl7rNdCNsREymcZ_nWw.png" /></figure><p>That infamous <strong>“your location has been shared 5,398 times”</strong> stat comes from a 2015 CHI field study by CMU and Notre Dame researchers. They wanted to know something very practical which is giving people a privacy dashboard is nice, but does it actually get used? And if people ignore it, can a nudge wake them back up and realize?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/362/1*Yc5djHyWzyKiMsugnT-cCg.png" /><figcaption>Screenshot of a privacy nudge for location (left) shown to a participant during our study. Nudge content is based on participant’s apps and their access to location information. “Let me change my settings” opens AppOps. “Show me more before I make changes” opens the detailed report (right). “Keep Sharing my location” closes the nudge</figcaption></figure><p>They recruited <strong>23 </strong>Android users and ran a <strong>22-day</strong> study using AppOps, an early Android permission manager.</p><ul><li><strong><em>Week 1:</em></strong> just watched what people did with zero intervention, to get a baseline.</li><li><strong><em>Week 2: </em></strong>gave people access to AppOps and let them explore on their own.</li><li><strong><em>Week 3:</em></strong> kept AppOps on, but added daily nudges like “Your location has been shared<strong> X</strong> times with these apps,” with three response buttons: open settings, see more detail, or keep sharing.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/912/1*Dmy7pRvqXq4cPU5OQ56OUw.png" /><figcaption>Timeline of participants’ interactions with AppOps and the privacy nudges during phase 2 and phase 3. In the X-axis, each column represents a day during phase 2 (7 days) &amp; phase 3 (8 days)</figcaption></figure><p>I noticed something interesting which was just handing someone a privacy dashboard worked, once. <strong>22 </strong>of <strong>23</strong> people opened it and reviewed permissions in week 2. But interest dropped off fast, down to single digits by <strong>day 10</strong>. A tool existing isn’t the same as a tool getting used repeatedly. People check once, feel like they have <strong>“handled it,”</strong> and just move on which is concerning.</p><p>The surprising part to me was how effective the nudges were, specifically because of the number, not the design. In interviews, participants kept zeroing in on the raw access count. One person described feeling <strong>“followed by my own phone”</strong> after seeing an app had grabbed their location over <strong>4,000 </strong>times. Nobody mentioned the button layout or the color scheme. The number did all the emotional work. That’s a bigger lesson than it sounds like the nudge wasn’t persuasive because it was well-designed, it was persuasive because it was honest and specific.</p><p>A few things worth noticing for anyone designing a notification system:</p><ul><li><strong>95%</strong> of people reviewed their permissions again after a nudge, and <strong>58%</strong> went further and restricted something</li><li>Most restrictions happened on apps people had stopped using, or apps whose access made no obvious sense (why does a weather app need your location thousands of times a day right?) Kind of suspicious.</li><li>Repeating the same nudge a few days later was noticeably less effective. The second time someone saw the same type of alert, they were less likely to act on it.</li><li>A group of people dismissed nudges not because they didn’t care, but because the timing was bad one participant got a full-screen nudge mid-run that blocked their running app</li></ul><p>The open question I keep circling this was <strong>22 days</strong>. Real apps, real notifications, real permission requests happen for years, not three weeks. I would want to know whether these nudges still work after someone’s seen thousands of them, or whether the “being followed by my own phone” reaction fades the same way ignoring cookie banners became automatic. My first move would be to look for any longer-term field studies on notification fatigue specifically for privacy nudges, since that’s the exact gap this paper flags as unsolved.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a61d1c46a354" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Stop Trusting the Color, Read the Feedback]]></title>
            <link>https://medium.com/@sufyankr41/stop-trusting-the-color-read-the-feedback-9bd564837871?source=rss-48acd99e7c72------2</link>
            <guid isPermaLink="false">https://medium.com/p/9bd564837871</guid>
            <category><![CDATA[data-driven]]></category>
            <category><![CDATA[password-meter]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[data-protection]]></category>
            <category><![CDATA[passwords]]></category>
            <dc:creator><![CDATA[Sufyan Muhammad]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 13:11:35 GMT</pubDate>
            <atom:updated>2026-10-01T13:32:00.859Z</atom:updated>
            <content:encoded><![CDATA[<p>Every signup form has one: that little bar under the password field that turns red, then yellow, then green. Most of us have learned to just add a number and an exclamation mark until it turns green, then move on. A 2017 CHI paper by researchers at CMU and the University of Chicago asked a blunt question, does that bar actually make passwords stronger, or are we just gaming a meter that doesn’t know what it’s talking about?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RHX7OtOMrXZ1nR-edgkrfw.png" /></figure><p>Here’s roughly what they built and tested:</p><ul><li>A password meter that didn’t just show a bar, but used a neural network plus 21 heuristics to actually understand <em>why</em> a password was weak (dictionary words, predictable patterns like “1qaz2wsx,” digits always stuck at the end)</li><li>Based on that, it gave specific text feedback not “weak,” but something like “try moving your digits away from the end” or “avoid common keyboard patterns”</li><li>It even suggested a tweaked version of your own password, swapping a letter’s case or sliding a symbol into the middle, instead of asking you to start over</li><li>They tested all this on 4,509 people, under two different password rules (a lax one requiring 8+ characters, and a strict one requiring <strong>12+</strong> characters with mixed character types)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/477/1*YVALitONu3BJMkceCJ_dNg.png" /><figcaption>Guessability of medium-stringency passwords created without any feedback (“None”), with only a colored bar (“Bar”), and with both a colored bar and text feedback (“Std”)</figcaption></figure><p>I noticed one more important detail is how much the detailed text feedback outperformed the bar on its own. Under the easier password policy, people who got full explanations created passwords that were <strong>44%</strong> harder to guess than people who got no feedback at all, and the bar by itself barely moved the needle compared to that. It wasn’t the color changing that helped. It was being told, specifically, what was wrong.</p><p>The ceiling effect under the stricter policy took me off guard. Once people were already forced into <strong>12+</strong> characters with mixed types, the extra text feedback barely helped at all. Security effort has diminishing returns once the baseline rule is already doing the heavy lifting. That reframes how I think about combining interventions a strict policy and a smart meter aren’t additive, they’re somewhat redundant once one is doing its job.</p><p>A few other things worth pulling out for anyone designing onboarding flows:</p><ul><li>Memorability didn’t suffer. People created meaningfully stronger passwords with full feedback, and recall rates a few days later stayed about the same</li><li>Most people never clicked the “(Why?)” explanation link only about 1% did so the value was in the glance-able feedback itself, not the deep explanation</li><li>Suggested password rewrites helped some people, but a chunk of users straight up didn’t trust a computer tweaking their password, which is a trust problem more than a UX one</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g4J6qwqW1tEMFNP6gsfhDA.png" /></figure><p>The open question I keep coming back to is that all measured for a single password created once, in a lab-style study. Real people reuse passwords, get fatigued by the fifteenth account they’ve had to set up that month, and stop reading feedback they’ve seen a hundred times before. I’d want to know whether this kind of meter still works after someone’s seen it fifty times, or whether the novelty is doing some of the lifting here. Worth digging into any follow-up field studies that tracked the same users over months instead of a single session.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9bd564837871" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Privacy Policy Words That Quietly Lie to You]]></title>
            <link>https://medium.com/@sufyankr41/the-privacy-policy-words-that-quietly-lie-to-you-463f22d0ee12?source=rss-48acd99e7c72------2</link>
            <guid isPermaLink="false">https://medium.com/p/463f22d0ee12</guid>
            <category><![CDATA[privacy-policy]]></category>
            <category><![CDATA[understanding]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[whatsapp]]></category>
            <category><![CDATA[technical-terms]]></category>
            <dc:creator><![CDATA[Sufyan Muhammad]]></dc:creator>
            <pubDate>Tue, 29 Sep 2026 07:36:23 GMT</pubDate>
            <atom:updated>2026-10-01T12:30:28.965Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3OjEfDOHl32kELgjx4k1JQ.png" /><figcaption>(Privacy is a myth)</figcaption></figure><p>Privacy policies tend to open with words that sound familiar: metadata, encryption, aggregated information, local storage. Most of us skim right past them, assuming we know what they mean. Researchers at <strong>Wellesley College and Pomona College</strong> decided to test that assumption instead of taking it for granted, and what they found is worth stopping on.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/885/1*3rA07rDQw4ptWEFga2cq9Q.png" /><figcaption>User accuracy at defining technical terms that appear in privacy policies</figcaption></figure><blockquote><strong>The Words We Think We Understand</strong></blockquote><p>They pulled <strong>57</strong> technical terms from real privacy policies and asked nearly 800 people to explain each one. The results were not promising. The average score didn’t even reach <strong>40%</strong>, and on many terms less than half of respondents got it right. About <strong>28%</strong> believed <strong>“device fingerprinting”</strong> meant using your actual fingerprint to unlock a phone, which is absurd. Around one in three figured <strong>“local storage” </strong>meant your data physically sits somewhere close to you.</p><blockquote><strong>When Boring Words Build False Trust</strong></blockquote><p>The mistakes themselves were what stuck with me. I expected people to misread a term and panic over nothing, and that was not always the case. Sometimes the error ran the other way. <strong>“We collect metadata”</strong> sounds like a boring marker, so people relaxed. They didn’t realize metadata can reveal when you talked, with whom, and for how long. Wording can do more than scare people. It can build trust that hasn’t been earned.</p><p>If I were building a tool around this, that’s the warning I would personally take from it. The lesson isn’t “make things sound frightening or scary.” It’s that plain, comforting language that can make people feel safe when they are really not.</p><blockquote><strong>Why Encryption Understanding Isn’t Fixed</strong></blockquote><p>The number that surprised me most was about end-to-end encryption. Only around <strong>17%</strong> said standard encryption was safe, while more than <strong>70%</strong> could explain what E2E actually does to protect people. Honestly, I did not expect that, especially after reading an older interview study on WhatsApp users, where almost everyone seemed confused about encryption.</p><p>The timing probably explains it, as this survey ran in 2021, well after a steady stream of news stories asking whether WhatsApp was really private. The people in the earlier study were likely hearing the concept for the first time. Understanding shifts as public conversation about a topic grows, so an explanation that flops today might land later, once enough people have heard it and passed it along.</p><blockquote><strong>What Surveys Can’t Tell Us And Makes Us Question?</strong></blockquote><p>One thing keeps revolving around my mind. The paper suggests different terminology shapes of how people feel and what they say they will put up with. But saying you’d accept something in a survey is different from tapping <strong>“I agree”</strong> in an app you actually want to use. In real life, most people just want to click through onboarding and get going. So I’d like to see field studies rather than surveys alone, such as <strong>tracking real opt-out rates</strong> after a clause is rewritten in plain language and comparing them against the technical version.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=463f22d0ee12" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Encryption Isn’t the Problem, Understanding It Is]]></title>
            <link>https://medium.com/@sufyankr41/encryption-isnt-the-problem-understanding-it-is-9e7e1e03b939?source=rss-48acd99e7c72------2</link>
            <guid isPermaLink="false">https://medium.com/p/9e7e1e03b939</guid>
            <category><![CDATA[communication-tools]]></category>
            <category><![CDATA[encryption]]></category>
            <category><![CDATA[security]]></category>
            <dc:creator><![CDATA[Sufyan Muhammad]]></dc:creator>
            <pubDate>Fri, 25 Sep 2026 11:29:20 GMT</pubDate>
            <atom:updated>2026-09-25T11:29:20.521Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xVngAR2GC-_F3BuVQYFU9Q.png" /></figure><p>“Obstacles to the Adoption of Secure Communication Tools,” a 2017 study by UCL, the University of Bonn, Stanford, and the EFF, was the paper I read this week, in which 60 people were interviewed by the researchers to find out how users actually use programs like Skype, Telegram, WhatsApp, Signal, and iMessage, as well as what they think goes on behind the scenes before and after they do any activity.</p><blockquote>The Main Finding</blockquote><p>Usability isn’t what keeps people off secure messaging apps. Understanding is. Most participants didn’t switch tools because their contacts weren’t there, they judged <strong>“security”</strong> using signals that have nothing to do with security and almost nobody actually understood what end-to-end encryption does. Many believed encryption gets broken eventually anyway, so there wasn’t much point trusting it. Even the EFF’s own Secure Messaging Scorecard, built specifically to help non-experts pick safer tools, didn’t change a single person’s ranking, because the criteria on it (forward secrecy, open-source code, fingerprint verification) meant nothing without a concrete threat to attach them to. Some participants even read “open-source” as a bad sign, assuming visibility means vulnerability. The paper’s conclusion is blunt as I would say, instead of building new niche secure tools and hoping people migrate, security researchers should put their effort into securing the platforms people already use by the billions.</p><blockquote><strong>Interesting Part</strong></blockquote><p>What struck me most is that usability wasn’t really the problem. Trust was. Mental models were. I had assumed, like most of the security world seems to, that if you make encryption invisible and effortless, people will just use secure tools. But this paper shows people stick with tools they know are less secure, on purpose, because their contacts aren’t anywhere else. Convenience wins every time.</p><p>It’s relevant to Indonesia in an obvious way WhatsApp owns messaging there for the same reasons like how it works across devices, it’s cheap on data, everyone’s already on it. So if you wanted to remind people about more secure messaging, “build a better signal” isn’t the move. The better bet is making sure the encryption already inside WhatsApp is something people actually trust and understand, and putting effort into digital literacy around what encryption is even protecting against.</p><blockquote><strong>Numbers That Surprise</strong></blockquote><p>The surprising part to me was how deep the confusion went, even with educated participants. People called encryption “binary language.” Some thought using two passwords counts as “double encryption.” Others figured open-source code makes something less secure, since now attackers can see how it works. Out of 60 people, only <strong>three</strong> understood things like forward secrecy or digital signatures and all three were security experts by trade.</p><p>That redefined something for me that security awareness campaigns don’t fail because people are lazy. They fail because words like “E2E encryption” or “fingerprint verification” mean nothing without a real threat attached to them. For something like this in Indonesia school programs, government campaigns, fintech onboarding the fix probably isn’t better vocabulary. It’s skipping the vocabulary. “If your phone gets stolen, here’s what this actually stops” lands. “This app uses AES-256 encryption” doesn’t.</p><blockquote><strong>The Arising Question</strong></blockquote><p>The question that I keep circling my mind around is if better wording and scorecards don’t fix the misunderstanding, what does? The paper doesn’t pretend to know the authors admit completely that rewriting the EFF scorecard won’t touch the underlying mental models. I want to know this because it’s the kind of thing you could actually test. If I were building a fintech app or a government digital ID system, I’d want to know whether small in-app warnings at the moment of risk beat generic literacy content, or whether people trust a warning from family more than one from an app. The first step would be digging into newer HCI security papers that test interventions instead of just documenting the problem, and looking at how WhatsApp’s own “this chat is encrypted” banner has delayed under study it’s the closest real version of this experiment already running.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9e7e1e03b939" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>