Skip to main content

Differentially-Private Synthetic Dataset Release for Machine Learning and Clustering

Researcher: Avrim Blum

Research Area: Privacy Protection

Cross Cutting Thrusts: Usable Privacy and Security

Abstract

Scope: This project aims to advance technology for privacy-preserving synthetic dataset release in the demanding differential privacy model, as well as developing privacy formulations that extend this model with additional desirable properties for distributed databases in which each portion has its own owner.

Outcomes: New algorithms developed for preserving differential privacy as well as an improved understanding of the types of “usefulness” guarantees that can be provided for synthetic dataset construction in the differential privacy framework.