C4D Researchers bring home prizes from privacy-enhancing technologies conference

C4D researchers at the PETS conference.

Our lab’s cutting-edge research into online systems was well-represented at this year’s Privacy Enhanced Technology Symposium (PETS), hosted at George Washington University in D.C. In fact, two of our researchers took home prizes, for best student paper and best poster! The symposium, which had over 300 people in attendance, brings together interdisciplinary experts in the field of privacy-enhancing technologies and publishes a well-regarded journal that is openly available on their website.

C4D researchers Julia Kieserman and Cat Mai presented papers published in this year’s journal, which had an overall acceptance rate of 26%. 

Online Trackers are Not Created Equal

Julia’s paper about personal data collected by Meta and Google on sensitive health and finance websites won the Andreas Pfitzmann Best Student Paper Award. This research shows how Google and Meta encourage website administrators to not only install tracking scripts on their websites, but configure them to automatically collect users’ Personally Identifying Information (PII). Read our recent blog post describing our key takeaways from this work and check out the full paper.

Understanding the Perils of YouTube's Ad Privacy Settings on Safety

Cat presented her new paper, “More and Scammier Ads: The Perils of YouTube’s Ad Privacy Settings” and won the Best Poster Award. In what we believe is the first study of its kind, Cat’s research demonstrated that enabling the strongest privacy settings on Google netted users vastly more ads – particularly predatory ones – on YouTube. The results raise a number of concerns and questions about Google’s ad transparency as well as ad delivery processes. 

As the paper lays out, we found that disabling ad personalization on YouTube leads to a worse viewing experience for privacy-conscious users. Those with the most private settings saw 30% more pre-roll ads and more concerningly, 2.69 times more predatory ads compared to less cautious viewers. Those watching conspiracy videos with their privacy settings maximized were hit with the most scam ads: They were 4.45 times as likely to see a predatory pre-roll ad than someone using the most lax privacy settings, the researchers found. 

Why is this happening? We hypothesize that legitimate advertisers pay more for viewers who share data, so when users withhold that data, Google fills the gap with cheaper, lower-quality ads which tend to be more predatory.

Our experiments also show that YouTube’s ad explanations are incomplete and fail to reveal why a specific ad was shown, and often are identical across very different ad types. We suspect they only reflect advertiser-defined targeting, omitting other delivery signals, which may violate Article 26 of the EU Digital Services Act in Ireland. Our findings raise serious questions about Google’s ability – and willingness – to curb these ads. You can read the full paper, which was supported by the National Science Foundation, here.