We introduce new differentially private (DP) mechanisms for gradient-based machine learning (ML) with multiple passes (epochs) over a dataset, substantially improving the achievable privacy-utility-computation tradeoffs. We formalize the problem of DP mechanisms for adaptive streams with multiple participations and introduce a non-trivial extension of online matrix factorization DP mechanisms to our setting. This includes establishing the necessary theory for sensitivity calculations and efficient computation of optimal matrices. For some applications like >10,000 SGD steps, applying these optimal techniques becomes computationally expensive. We thus design an efficient Fourier-transform-based mechanism with only a minor utility loss. Extensive empirical evaluation on both example-level DP for image classification and user-level DP for language modeling demonstrate substantial improvements over all previous methods, including the widely-used DP-SGD . Though our primary application is to ML, our main DP results are applicable to arbitrary linear queries and hence may have much broader applicability.
@article{arxiv.2211.06530,
title = {Multi-Epoch Matrix Factorization Mechanisms for Private Machine Learning},
author = {Christopher A. Choquette-Choo and H. Brendan McMahan and Keith Rush and Abhradeep Thakurta},
journal= {arXiv preprint arXiv:2211.06530},
year = {2023}
}
Comments
9 pages main-text, 3 figures. 40 pages with 13 figures total