Certified private data release for sparse Lipschitz functions

TL;DR

This paper introduces a differentially private data release algorithm tailored for sparse Lipschitz functions, providing fast utility loss rates and a method to certify utility loss for various algorithms, enhancing privacy-preserving machine learning.

Contribution

The work presents a novel algorithm with fast utility loss rates for sparse Lipschitz queries and a certification method applicable to many algorithms.

Findings

Achieves fast utility loss rates for sparse Lipschitz functions

Provides a certification method for utility loss in private algorithms

Enhances privacy guarantees in machine learning data releases

Abstract

As machine learning has become more relevant for everyday applications, a natural requirement is the protection of the privacy of the training data. When the relevant learning questions are unknown in advance, or hyper-parameter tuning plays a central role, one solution is to release a differentially private synthetic data set that leads to similar conclusions as the original training data. In this work, we introduce an algorithm that enjoys fast rates for the utility loss for sparse Lipschitz queries. Furthermore, we show how to obtain a certificate for the utility loss for a large class of algorithms.