On the Fairness of Privacy Protection: Measuring and Mitigating the Disparity of Group Privacy Risks for Differentially Private Machine Learning
This addresses fairness issues in privacy protection for differentially private machine learning, offering a more reliable assessment and mitigation of group disparities, though it is incremental as it builds on existing DPML methods.
The paper tackled the problem of unfair privacy protection across groups in differentially private machine learning by introducing a novel membership inference game to measure worst-case group privacy risks more efficiently and enhancing DP-SGD with an adaptive group-specific gradient clipping strategy, which reduced disparity in group privacy risks as confirmed by experiments.
While significant progress has been made in conventional fairness-aware machine learning (ML) and differentially private ML (DPML), the fairness of privacy protection across groups remains underexplored. Existing studies have proposed methods to assess group privacy risks, but these are based on the average-case privacy risks of data records. Such approaches may underestimate the group privacy risks, thereby potentially underestimating the disparity across group privacy risks. Moreover, the current method for assessing the worst-case privacy risks of data records is time-consuming, limiting their practical applicability. To address these limitations, we introduce a novel membership inference game that can efficiently audit the approximate worst-case privacy risks of data records. Experimental results demonstrate that our method provides a more stringent measurement of group privacy risks, yielding a reliable assessment of the disparity in group privacy risks. Furthermore, to promote privacy protection fairness in DPML, we enhance the standard DP-SGD algorithm with an adaptive group-specific gradient clipping strategy, inspired by the design of canaries in differential privacy auditing studies. Extensive experiments confirm that our algorithm effectively reduces the disparity in group privacy risks, thereby enhancing the fairness of privacy protection in DPML.