Overview
Explore the concept of differential privacy in this distinguished lecture by Cynthia Dwork from Harvard University's Radcliffe Institute for Advanced Study. Delve into the mechanisms, relaxations, and geometric aspects of differential privacy, understanding its importance and current applications. Examine case studies involving Facebook and the US Census Bureau, and investigate challenges such as overly accurate estimates and database reconstruction. Learn about differentially private synthetic data, hardness results, and future directions in the field. Engage with the Feinberg problem and sample-and-aggregate approaches, gaining insights into this crucial aspect of data privacy and its implications for statistical analysis and societal impact.
Syllabus
Introduction
Differential Privacy
Mechanisms
Relaxations
L1 Sensitivity
Geometry
Why I love it
Where are we today
Facebook
Census Bureau
Activity
Overly Accurate Estimates
US Census
Database Reconstruction Theorem
Statistics
Challenges
Differentially Private Synthetic Data
Blum Liggett Roth
Hardness Results
Future Directions
Steve Feinberg
Two General Approaches
The Feinberg Problem
Sample and Aggregate
The Problem
Questions
Taught by
TheIACR