Bounding Training Data Reconstruction in Private (Deep) Learning

TL;DR

This paper establishes formal semantic guarantees for differential privacy mechanisms, demonstrating their effectiveness in preventing training data reconstruction attacks in machine learning models.

Contribution

It introduces the first semantic guarantees for DP against data reconstruction attacks, analyzing Renyi differential privacy and Fisher information leakage as protective measures.

Findings

Renyi differential privacy provides strong semantic protection.

Fisher information leakage effectively limits data reconstruction risks.

Both methods enhance understanding of privacy guarantees in ML models.

Abstract

Differential privacy is widely accepted as the de facto method for preventing data leakage in ML, and conventional wisdom suggests that it offers strong protection against privacy attacks. However, existing semantic guarantees for DP focus on membership inference, which may overestimate the adversary's capabilities and is not applicable when membership status itself is non-sensitive. In this paper, we derive the first semantic guarantees for DP mechanisms against training data reconstruction attacks under a formal threat model. We show that two distinct privacy accounting methods -- Renyi differential privacy and Fisher information leakage -- both offer strong semantic protection against data reconstruction attacks.