The Impact of Differential Privacy on Group Disparity Mitigation

TL;DR

This paper investigates how differential privacy affects fairness across various tasks, revealing that privacy can both hinder and help fairness depending on the training approach, and interprets privacy as a form of regularization.

Contribution

It provides a comprehensive evaluation of differential privacy's impact on fairness across multiple tasks and introduces a reinterpretation of privacy as regularization.

Findings

Differential privacy increases group disparities in baseline models.

In robust training, differential privacy reduces group disparities.

Privacy acts as a regularizer influencing fairness outcomes.

Abstract

The performance cost of differential privacy has, for some applications, been shown to be higher for minority groups; fairness, conversely, has been shown to disproportionally compromise the privacy of members of such groups. Most work in this area has been restricted to computer vision and risk assessment. In this paper, we evaluate the impact of differential privacy on fairness across four tasks, focusing on how attempts to mitigate privacy violations and between-group performance differences interact: Does privacy inhibit attempts to ensure fairness? To this end, we train $(\varepsilon,\delta)$-differentially private models with empirical risk minimization and group distributionally robust training objectives. Consistent with previous findings, we find that differential privacy increases between-group performance differences in the baseline setting; but more interestingly, differential privacy reduces between-group performance differences in the robust setting. We explain this by reinterpreting differential privacy as regularization.