Optimal Privacy-Constrained Mechanisms
summary
Work introduces a Bayesian privacy-loss metric measured by KL divergence (equivalently mutual information) and enforces an exogenous privacy budget kappa on mechanisms; solves a seller’s screening problem with one buyer under quadratic costs. Optimal mechanisms use an interval partition of types (privacy constraint binds; very tight privacy reduces to two intervals), the number of intervals jumps discretely as kappa relaxes (visible kinks in profit), full privacy yields posted-price buyer surplus maximization, no privacy maximizes total welfare.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:06Differential Privacy
- 1:38What if
- 2:47General Mechanism
- 3:43Privacy Laws
- 4:04Exercise
- 4:44Discussion
- 10:02Application
- 13:27Mean result
- 13:53Properties
- 23:15Consumer Surplus Total Welfare
- 23:59Summary
- 24:56Conclusion
related talks
01h 05mMarkets and Incentives for Transacting on (Personal) Information, Part III
44mPrivate Meta-Selection
35mPrivacy amplification by random allocation
39mPufferfish Privacy Mechanisms for Correlated Data
44mAccuracy Disparities and Social Choices in the Design of Privacy Mechanisms
16mBalancing Privacy: Leveraging Privacy Budgets for Privacy-Enhancing Technologies
26mLocally Private Histograms in All Privacy Regimes
43mPufferfish Privacy Mechanisms for Correlated Data