Subsampled Renyi Differential Privacy and Analytical Moments Accountant

[edit]

Yu-Xiang Wang, Borja Balle, Shiva Prasad Kasiviswanathan ;
Proceedings of Machine Learning Research, PMLR 89:1226-1235, 2019.

Abstract

We study the problem of subsampling in differential privacy (DP), a question that is the centerpiece behind many successful differentially private machine learning algorithms. Specifically, we provide a tight upper bound on the Renyi Differential Privacy (RDP) [Mironov 2017] parameters for algorithms that: (1) subsample the dataset, and then (2) applies a randomized mechanism M to the subsample, in terms of the RDP parameters of M and the subsampling probability parameter. Our results generalize the moments accounting technique, developed by [Abadi et al. 2016] for the Gaussian mechanism, to any subsampled RDP mechanism.

Related Material