Predicate Singling Out: Towards Formalizing the GDPR Notion of Singling Out
Talkat Data Privacy: From Foundations to Applications (Simons Institute workshop, 2019) · 2019-03-04..03-08·
summary
Formalizes GDPR “singling out” as Predicate Singling Out (PSO): an adversary outputs a predicate whose weight (probability an individual satisfies it) is tiny yet matches exactly one row. Shows a trivial baseline peaks near 37% baseline at weight 1/n, uses hashing/leftover hashing to avoid distribution knowledge, and proves the counting mechanism (publishing exact counts) is PSO‑secure while k‑anonymity fails PSO. Demonstrates a differencing attack that composes after ~log n queries so PSO doesn't compose, and that (ε,δ)-DP implies PSO security.
AI-written summary · about
chapters (5)
- 0:00Start
- 4:35Extract: singling out and the GDPR
- 13:51Formalize: predicate singling out
- 29:38Analyze: composition, k-anonymity, differential privacy
- 39:30Draw legal conclusions
related talks
01h 16mLaw and CS
18mProtecting User Privacy in Private Set Intersection: A Journey Toward Mitigating User
01h 14mStatistical Inference and Privacy, Part III
19mPractical Considerations for Differential Privacy
26mAaron Roth, Associate Professor, University of Pennsylvania, NYC 2017
26mGeneralized Private Selection and Testing with High Confidence
14mEstablishing Privacy Metrics for Genomic Data Analysis
35mPrivacy amplification by random allocation