English

Differentially Private Image Classification by Learning Priors from Random Processes

Computer Vision and Pattern Recognition 2023-11-01 v2 Cryptography and Security Machine Learning Machine Learning

Abstract

In privacy-preserving machine learning, differentially private stochastic gradient descent (DP-SGD) performs worse than SGD due to per-sample gradient clipping and noise addition. A recent focus in private learning research is improving the performance of DP-SGD on private data by incorporating priors that are learned on real-world public data. In this work, we explore how we can improve the privacy-utility tradeoff of DP-SGD by learning priors from images generated by random processes and transferring these priors to private data. We propose DP-RandP, a three-phase approach. We attain new state-of-the-art accuracy when training from scratch on CIFAR10, CIFAR100, MedMNIST and ImageNet for a range of privacy budgets ε[1,8]\varepsilon \in [1, 8]. In particular, we improve the previous best reported accuracy on CIFAR10 from 60.6%60.6 \% to 72.3%72.3 \% for ε=1\varepsilon=1.

Keywords

Cite

@article{arxiv.2306.06076,
  title  = {Differentially Private Image Classification by Learning Priors from Random Processes},
  author = {Xinyu Tang and Ashwinee Panda and Vikash Sehwag and Prateek Mittal},
  journal= {arXiv preprint arXiv:2306.06076},
  year   = {2023}
}
R2 v1 2026-06-28T11:01:19.568Z