Private and Fair Machine Learning: Revisiting the Disparate Impact of Differentially Private SGD

2025-10-02

AI Summary

This study investigates the heterogeneous impact of differentially private stochastic gradient descent (DPSGD) on model utility and fairness under privacy constraints. We systematically evaluate accuracy and multiple fairness metrics across diverse tasks and datasets, performing comprehensive hyperparameter sweeps for both standard DPSGD and its Global-Adapt variant. Results show that direct hyperparameter optimization on private models—while insufficient to fundamentally mitigate unfairness—significantly improves the utility–fairness trade-off. In contrast, DPSGD-Global-Adapt exhibits high sensitivity to hyperparameters and poor robustness. Crucially, hyperparameter tuning itself incurs additional privacy leakage, necessitating explicit accounting within the overall privacy budget. To our knowledge, this is the first work to quantitatively characterize the triadic tension among privacy, utility, and fairness in DPSGD. Our findings provide empirical grounding and methodological cautions for trustworthy hyperparameter selection in privacy-preserving machine learning.

Technology Category

• Machine Learning: Privacy
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Computer Vision: Bias, Fairness & Privacy

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

Differential privacy (DP) is a prominent method for protecting information about individuals during data analysis. Training neural networks with differentially private stochastic gradient descent (DPSGD) influences the model's learning dynamics and, consequently, its output. This can affect the model's performance and fairness. While the majority of studies on the topic report a negative impact on fairness, it has recently been suggested that fairness levels comparable to non-private models can be achieved by optimizing hyperparameters for performance directly on differentially private models (rather than re-using hyperparameters from non-private models, as is common practice). In this work, we analyze the generalizability of this claim by 1) comparing the disparate impact of DPSGD on different performance metrics, and 2) analyzing it over a wide range of hyperparameter settings. We highlight that a disparate impact on one metric does not necessarily imply a disparate impact on another. Most importantly, we show that while optimizing hyperparameters directly on differentially private models does not mitigate the disparate impact of DPSGD reliably, it can still lead to improved utility-fairness trade-offs compared to re-using hyperparameters from non-private models. We stress, however, that any form of hyperparameter tuning entails additional privacy leakage, calling for careful considerations of how to balance privacy, utility and fairness. Finally, we extend our analyses to DPSGD-Global-Adapt, a variant of DPSGD designed to mitigate the disparate impact on accuracy, and conclude that this alternative may not be a robust solution with respect to hyperparameter choice.

Problem

• Analyzing disparate impact of differentially private SGD on fairness
• Evaluating hyperparameter optimization effects on utility-fairness trade-offs
• Assessing privacy leakage risks in balancing privacy, utility, fairness

Innovation

• Direct hyperparameter optimization on differentially private models
• Analyzing disparate impact across multiple performance metrics
• Evaluating utility-fairness trade-offs in private SGD variants