Subsampled Rényi Differential Privacy and Analytical Moments Accountant
Keyword(s):
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.
Keyword(s):
2020 ◽
Vol 2020
(1)
◽
pp. 103-125
Keyword(s):
2019 ◽
Vol 1
(2)
◽
pp. 78-80
Keyword(s):
2017 ◽
Vol 12
(1)
◽
pp. 21
◽
Keyword(s):
Keyword(s):
Keyword(s):