Subsampled Rényi Differential Privacy and Analytical Moments Accountant
Journal of Privacy and Confidentiality · 2020-06-11 · 10 (2)
DOI: 10.29012/jpc.723 ↗At a glance
Journal article in Journal of Privacy and Confidentiality, published online on 2020-06-11. Written by Yu-Xiang Wang and 2 co-authors. Cited 102 times so far (Crossref). The full text is freely available (open access, CC BY-NC-ND 4.0).
Abstract (original language)
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) apply 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.
Cite
APA
Wang, Y., Balle, B., & Kasiviswanathan, S. (2020). Subsampled Rényi Differential Privacy and Analytical Moments Accountant. Journal of Privacy and Confidentiality, 10(2). https://doi.org/10.29012/jpc.723
Vancouver
Wang Y, Balle B, Kasiviswanathan S. Subsampled Rényi Differential Privacy and Analytical Moments Accountant. Journal of Privacy and Confidentiality. 2020;10(2). doi:10.29012/jpc.723
BibTeX
@article{Wang2020,
author = {Wang, Yu-Xiang and Balle, Borja and Kasiviswanathan, Shiva},
title = {Subsampled Rényi Differential Privacy and Analytical Moments Accountant},
journal = {Journal of Privacy and Confidentiality},
year = {2020},
volume = {10},
number = {2},
doi = {10.29012/jpc.723}
}
EUVIMED