Subsampled R\'enyi Differential Privacy and Analytical Moments Accountant
Machine Learning
2018-12-05 v2 Cryptography and Security
Machine Learning
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 R\'enyi 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.
Keywords
Cite
@article{arxiv.1808.00087,
title = {Subsampled R\'enyi Differential Privacy and Analytical Moments Accountant},
author = {Yu-Xiang Wang and Borja Balle and Shiva Kasiviswanathan},
journal= {arXiv preprint arXiv:1808.00087},
year = {2018}
}