Global changes

The name of the framework has been globally changed from contact tracing to encounter notification. This is an appropriate change, since it does very little of what a traditional contact tracing system does; it just designed to provide notification of an encounter with an infected individual.

Bluetooth spec

• Broadcast packet include metadata, currently version number and transmit power. Metadata is encrypted so it can’t be used to fingerprint devices, can only be decoded when matching diagnosis key is found. Including the transmit power is important as it will allow better estimates of the distance between devices.
• Additional clarification about how the temporary mac addresses and the rolling proximity keys are rotated in tandem.

Cryptography spec

• AES is now used instead of HMAC<SHA256> for improved performance.
• Daily tracing key renamed temporary exposure key
• Rotation period for temporary exposure key parameterized, currently still one day, but could be changed (e.g., to 12 hours or 48 hours)
• temporary exposure key are randomly generated and the ones for the past 14 days are stored. Previously, computed from seed key and day number, which provided weaker privacy guarantees
• When a received Bluetooth message is matched against a diagnosis key, the times are allowed to be off by as much as two hours (although they have to be from the same ENIntervalNumber). Presumably this is intended for an encounter to be detected by two devices whose clocks are out of sync by as much as two hours. However, it also makes it easier for malicious attackers to re-broadcast bluetooth advertisements in relay or replay attacks.

API

• When a device scans a set of diagnosis keys, the framework used to return just the number of matched keys. It now also returns the number of days since the most recent of the matched keys.
• When the framework is asked to give details about the encountered diagnosis keys, it used to just return the date of the exposure and the duration (in 5 minute increments). It now says the duration is capped at 30 minutes (so you can’t tell the difference between a 30 minute exposure and a 90 minute expose) and it also discloses the closest/minimum observed signal strength attenuation: the transmit power minus the RSSI of the received message. Possible addition: It might be useful to include average signal strength, so a very close encounter for 30 minutes can be distinguished from a distant 30 minute encounter with a momentary close encounter.
• The method for getting details of the encountered keys allows filtering based on minimum duration and maximum signal strength attenuation. This doesn’t actually provide any new capabilities, since that data is included in the reported encounters and they could be filtered by the same parameters after they are given to the app.
• Methods added to reset temporary exposure key and to delete all data.
• Other minor changes to how framework would be invoked, but don’t seem to effect functional behavior.

Overall impressions

All of the changes are good, some of them are important changes that address problems, others minor. A number of minor glitches and ambiguities I had identified in the initial version of the framework have been resolved. The only change I don’t undertand is the need to allow the send/received time for packets to be mismatched by up to two hours.

I still have several concerns:

• How health organizations and individuals will gain confidence in the value of using the framework
• How effective it will be at detecting epidemiologically significant encounters (encounters where the disease could be transmitted).
• When people receive an alert of a notification, will they will have the understanding and confidence to take appropriate action.
• Assuming sufficient resources, I think traditional contact tracing for covid-19 involving public health workers will be critically important, perhaps more important than adoption of a Bluetooth encounter notification system. While it is important to be able to use the encounter notification framework in an independent way that maximizes privacy, I think there are some valuable synergies that could be found by integrating the two approaches, and the current API doesn’t accommodate that use case.

Our trip to Southeast Asia starts in Singapore and ends in Ho Chi Min City. We actually pass through Ho Chi Min City twice. We spend the first part of our trip on our own or with a guide, and then once we reach Laos, we transition to a Backroads bicycle trip that goes from Laos to Cambodia and then a river cruise along the Mekong River ending in Ho Cho Min City.

Our trip to Southeast Asia was originally published in Adventures with Bill and Lisa on Medium, where people are continuing the conversation by highlighting and responding to this story.

I have no official role in the hiring process, but I have been a strong advocate for undergraduate education and for hiring exceptional instructional faculty. I want to tell you a couple of reasons why I think this is a particularly exciting opportunity:

• In Fall 2018, the entire department will be moving into the new Iribe Center for Computer Science and Innovation. In addition to nice offices, research labs and community spaces, the Iribe Center will contain
• Six classrooms, ranging from 60 to 300 seats, all designed to facilitate collaborative and small group activity during class.
• A 5000 sq foot makerspace intended for independent undergraduate student projects, with equipment such as laser cutters, 3D printers, electronics, soldering irons, and a wood shop.
• There is a $200,000 fund available to support innovation in undergraduate education. This can be used for special equipment, professional development and conference trips for instructional faculty, and other special activities (it can’t be siphoned into paying for faculty or TA salaries).
• Qualified and interested individuals might be given responsibilities in the makespace or in education innovation and technology in lieu of some of the classroom responsibilities that is part of the typical workload for instructional faculty.

I feel confident that exceptional candidates will be enthusiastically received, and that the department is flexible enough to make attractive offers to recruit such people. If you want something that isn’t in the official job posting, ask for it. For example, if you want to be considered for a job title of clinical professor or professor of practice, rather than lecturer, or if you want to be allowed to allocate some time to research, ask for that.

Feel free to apply directly to the ad, or contact me if you have any questions.

Bill Pugh, Professor Emeritus of Computer Science at UMD

Note: The University of Maryland actively subscribes to a policy of equal employment opportunity, and will not discriminate against any employee because of race, age, sex, color, sexual orientation, physical or mental disability, religion, ancestry or national origin, marital status, genetic information, or political affiliation. Minorities and women are encouraged to apply.

I do a fair bit of investing in stocks, my capital gains vary widely over the years, and I use the tax rules to minimize my taxes. Looking through that lens, most of the excitement about Trump’s $916M tax loss seems to be missing the point.

Let’s say all of Trump’s income/losses came from investing in stocks, since I understand that. To generate a $918M net loss, he would have had to purchased the stock for more than $918M. To do that, he would have needed to acquire the money to do so, which would have been taxable income. So he would have already paid taxes on more than $918M in income.

Furthermore, if he is even a quarter as rich as he says he is now, he would have had to earn billions since 1995, which would have used up all of the tax loss, and he wouldn’t get 18 years of tax free income from that one loss.

So he was smart to use a net loss to reduce his income taxes. I’d be surprised if he didn’t.

Rather, the question is whether or not he was using shady strategies to inflate his net losses or hide his net income. Real estate offers plenty of possibilities here, but there are others that apply more generally. For example, if you have a huge loss on an investment, you incur a loss. But if the people who you owe money to decide that you aren’t able to make good on your loss and let you pay only a fraction of your loss, the debt that is forgiven counts as taxable income, offsetting the loss.

But there are strategies that allow you to effectively get a loss forgiven, but not have that debt forgiveness officially recognized, and thus not have it count as taxable income. One such strategy is “debt parking”. See John Hempton’s write up on it (key excerpt below). There are also ways to hide or postpone what would otherwise be a taxable gain.

In short, the key question is whether Trump’s losses/gains in his tax filings accurately reflect the losses/gains in his net worth. If they do, then he did nothing unethical and it would have been proper for him to have not paid taxes for a number of years, although he would seem to have been a lousy investor.

But my expectation is that he at least pushed the boundaries of legal ways to inflate losses and hide gains, if not exceeded them. And that is why he is hiding his tax returns.

http://brontecapital.blogspot.com/2016/10/some-comments-on-new-york-times-story.html

Suppose the debtor (in this case The Donald) is going to get his debt cancelled for (say) 1c in the dollar. When he gets the debt wiped out the debtor (ie The Donald) will have to report assessable income equal to the debt wiped out (in this case 99 percent of $916 million).

The alternative though is for the debtor to set up a dummy party. The dummy party might be his wife or children or some company or trust set up by them or more likely some completely opaque offshore trust.

And that dummy party goes and buys the debt for say 1.1 cents in the dollar. Then they just sit there.

They don’t force the debtor (ie The Donald) to repay. They don’t make a profit or loss on the debt. And because the debtor never has his debt forgiven he never gets the assessment on debt forgiveness and he gets to keep his NOLs even though the losses did not come out of his pocket.

I’d like to discuss what comes after this case.

Apple is almostly certainly looking at whether they can modify iOS so that updating the OS on a locked phone wipes the phone (see Matt Blaze’s tweet). One impact of doing so is that if a phone has been bricked so that it can’t be unlocked (e.g., error 53), updating the OS to unbrick it will wipe the phone, and you’ll be out of luck if the phone isn’t backed up. Apple could probably get this done by iOS 10.0, and I wouldn’t be surprised if it came out in the final version of iOS 9.3 (likely to be released next month).

If FBI wins their case against Apple, I expect Apple to get lots of requests to apply the same “update” to other phones where there is a government warrant. Both the FBI and district attornies from all over the US will be filing requests. Foreign governments will also be making demands, and I don’t see how Apple could refuse those requests and still retain access to those foreign markets.

And as soon as the FBI extracts the information from Farook’s phone, you know they will turn the phone over the NSA to analyze the OS Apple installed. Foreign governments will also be turning the phones that Apple “updated” for them over to their cybersecurity groups for analysis. There is a significant chance that these “updates” would get into the wild, and people other than Apple would be able to apply them.

So I think that being in this situation would be bad for both Apple and Apple users. Even if you think the FBI should have access to any information it has a warrant for, I’m sure everyone can think of dissidents in foreign counties that should be able to keep secrets from their governments.

So I’m pretty confident that Apple will remove the ability to install an OS update on a locked phone without wiping the phone.

What happens if the FBI wins its case against Apple and Apple also pushes out an OS update that makes it impossible for Apple to update a locked phone without wiping the phone.

An OS update from Apple wouldn’t effect any phones already in the possession of the FBI or district attornies. But it would impact future inquiries.

But I can’t see the FBI giving up on access they had won in the courts. The FBI would probably demand that Apple push out an update to all phones that removed that security feature. This would probably be very controversial, and require legistation rather than just a judicial decision. A lot of politicians might decide this is a “good” solution to the problem of “going dark”.

Also, could the FBI use the San Bernardino precedent to argue that with a court warrant for a specific phone, they could force Apple to set things up so that when that phone next installed an over-the-air OS update, it got a special version of the OS update that included a backdoor for the FBI?

But I really don’t know. This would be very uncertain and untested waters.

Bill Pugh