Keynote Presentation: Minding the (Privacy Research/Practice) Gap
summary
A keynote arguing that differential privacy repeatedly fails or succeeds on concrete engineering and business terms: HIPAA eligibility for Alexa ships in 2019, lasts 3.5 years then rolls back; a DP team building HIPAA‑grade deidentification gets disbanded; Meta rebuilds ad measurement using aggregation, DP, secure multi‑party computation and on‑device models and recovers; Google’s Privacy Sandbox shows 15–17% reidentification among 1,000 profiles and rising to 57% after 15 epochs, 75% after 30; recommended fixes are embedded researchers, deployment‑linked publications, and joint industry–academic labs.
AI-written summary · about
chapters (15)
- 0:00Privacy research vs practice gap
- 2:39Industry partnerships poll
- 3:43Alexa HIPAA success
- 6:13HIPAA rollback challenges
- 9:28Privacy tech market fit
- 11:17SOUPS research impacts
- 13:23Differential privacy from SOUPS
- 14:46DP implementation examples
- 16:34Meta’s ad measurement success
- 18:20Privacy Sandbox failures
- 21:31Tumult Labs DP struggles
- 23:16DP product market fit struggles
- 23:58Call to action for researchers
- 25:44Research-industry collaboration
- 28:11Q&A
related talks
19mPractical Considerations for Differential Privacy
43mPanel: Privacy Design Patterns for AI Systems: Threats and Protections
45mMaking Privacy Technology Accessible: Benchmarks and Platforms
26mAaron Roth, Associate Professor, University of Pennsylvania, NYC 2017
44mHybrid Legal-Technical Concepts of Privacy
28mAdvertising, algorithms and Privacy: recent opensource developments
28mPrivacy Engineering Networking Break AMA
42mPrivacy as a Brand Differentiator