OK, I'm super late for this thread, but I promised — here are a few selected talks from this year's @PET_Symposium that I particularly liked ✨

Disclaimer: I only watched a small portion of talks, so I'm sure that this list is missing lots of great work. Let's start ⤵️
"The Power of the Hybrid Model for Mean Estimation" made me discover a new model halfway between local/central differential privacy.

For some problems like mean estimation, it naturally leads to significant utility gains.

https://petsymposium.org/2020/files/papers/issue4/popets-2020-0060.pdf
"Automatic Discovery of Privacy-Utility Pareto Fronts" proposes a clever way of computing privacy/utility trade-off graphs for differentially private ML models, very costly to train.

Its best student paper award is well-deserved!

https://petsymposium.org/2020/files/papers/issue4/popets-2020-0058.pdf
"The Price is (not) Right: Comparing Privacy in Free and Paid Apps" compares the privacy of paid/free apps in Android.

Besides the usual (many) creepy findings, it finds that paying doesn't really gets you much more privacy…

https://petsymposium.org/2020/files/papers/issue3/popets-2020-0050.pdf
"Angel or Devil? A Privacy Study of Mobile Parental Control Apps" looks at, well, parental control apps.

I dislike them in general, but I didn't expect that level of carelessness — data sent in the clear, trackers, no consent…

https://petsymposium.org/2020/files/papers/issue2/popets-2020-0029.pdf
"Averaging Attacks on Bounded Noise-based Disclosure Control Algorithms" breaks yet another bad "anonymization" algorithm with no formal guarantees.

Not surprising, sad that it's still necessary, very solid work regardless.

https://petsymposium.org/2020/files/papers/issue2/popets-2020-0031.pdf
"The TV is Smart and Full of Trackers: Measuring Smart TV Advertising and Tracking" does exactly what it says on the tin.

Tons of creepy findings of course. It's interesting how different the ecosystem is than on web/mobile.

https://petsymposium.org/2020/files/papers/issue2/popets-2020-0021.pdf
"Inferring Tracker-Advertiser Relationships in the Online Advertising Ecosystem using Header Bidding" investigates a flavor of real-time bidding, finds a clever way of reverse-engineering commercial relationships between trackers.


https://petsymposium.org/2020/files/papers/issue1/popets-2020-0005.pdf
"Differentially-Private Multi-Party Sketching for Large-Scale Statistics" combines crypto and DP to count unique users between untrusted actors.

I already mentioned this before — this is super clever, practical, impactful work.

https://petsymposium.org/2020/files/papers/issue3/popets-2020-0047.pdf
"An Analysis of the Current State of the Consumer Credit Reporting System in China" is an absolutely fascinating deep dive into how "social credit" works in China.

Extremely important work, on a topic often overlooked.
No recording, unfortunately.
https://petsymposium.org/2020/files/papers/issue4/popets-2020-0062.pdf
"A Privacy-Focused Systematic Analysis of Online Status Indicators" looks at these little things in online services, that tell your contacts whether you're online.

Excellent usability findings, thought-provoking recommendations.

https://petsymposium.org/2020/files/papers/issue3/popets-2020-0057.pdf
"No boundaries: data exfiltration by third parties embedded on web pages" is about creepy web tracker practices, like DOM exfiltration or login theft.

The impactful work is from 2017/2018 so the talk is a fascinating retrospective.

https://petsymposium.org/2020/files/papers/issue4/popets-2020-0068.pdf
"The Best of Both Worlds: Mitigating Trade-offs Between Accuracy and User Burden in Capturing Mobile App Privacy Preferences" proposes a very interesting approach to automated privacy assistants, that choose permissions for you.


https://petsymposium.org/2020/files/papers/issue1/popets-2020-0011.pdf
"ML Privacy Meter: Aiding regulatory compliance by quantifying the privacy risks of ML" proposes an automated way of determining whether a machine learning model can leak private training data.

It comes with a shiny open-source library to do exactly that!
"Probably private protocols" is quite thought-provoking — it proposes a class of private protocols in which you make additional non-collusion assumptions to gain performance, but still make sure everything doesn't break if the assumptions are wrong.
Still with me? I hope you won't mind me plugging my own work :D

"SoK: Differential Privacies" is a survey work on the many, many variants & extensions of differential privacy. The talk is accessible even if you don't know DP well.

https://petsymposium.org/2020/files/papers/issue2/popets-2020-0028.pdf
"Differentially Private SQL with Bounded User Contribution" presents our work to build a privacy-preserving query engine used by many people at Google.

Making DP tooling actually usable comes with lots of interesting challenges =)

https://petsymposium.org/2020/files/papers/issue2/popets-2020-0025.pdf
Finally, my favorite talk was probably the HotPETS keynote, "Privacy Threats in Intimate Relationships". Immensely thoughtful, fascinating, thought-provoking.

Honestly, this should be required watching if you build tech consumer products.
You can follow @TedOnPrivacy.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled:
Those who think consciousness raising & movement building are disconnected, or even a distraction from electoral politics have not paid attention to the history of social change.In this thread, we
Read more
The 2020s will be known as the Remote Work decadeA few predictions of what is likely to emerge[ a thread ] Third Space: Office and Working from Home will
Read more
1/29: Have you ever had a concept explained to you that helps frame complex issues you’ve been wrestling with and opens your eyes to new possibilities? A concept that I
Read more
By continuing to use the site, you are consenting to the use of cookies as explained in our Cookie Policy to improve your experience.