Performances of Symmetric Loss for Private Data from Exponential Mechanism

TL;DR

This paper investigates the robustness of symmetric loss in private learning using the exponential mechanism, providing theoretical insights, practical guidance on privacy budgets, and experimental validation on CIFAR-10.

Contribution

It offers a theoretical analysis of exponential mechanism properties with symmetric loss and proposes numerical guidance for privacy budgets in private learning.

Findings

Symmetric loss shows robustness in private data learning.

Numerical guidance improves privacy-utility trade-offs.

Experimental results validate theoretical insights on CIFAR-10.

Abstract

This study explores the robustness of learning by symmetric loss on private data. Specifically, we leverage exponential mechanism (EM) on private labels. First, we theoretically re-discussed properties of EM when it is used for private learning with symmetric loss. Then, we propose numerical guidance of privacy budgets corresponding to different data scales and utility guarantees. Further, we conducted experiments on the CIFAR-10 dataset to present the traits of symmetric loss. Since EM is a more generic differential privacy (DP) technique, it being robust has the potential for it to be generalized, and to make other DP techniques more robust.