Optimal Differentially Private Model Training with Public Data

arXiv:2306.15056v2 Announce Type: replace-cross
Abstract: Differential privacy (DP) ensures that training a machine learning model does not leak private data. In practice, we may have access to auxiliary public data that is free of privacy concerns. In this work, we assume access to a given amount of public data and settle the following fundamental open questions: 1. What is the optimal (worst-case) error of a DP model trained over a private data set while having access to side public data? …

access arxiv cs.cr cs.lg data differential privacy free leak machine machine learning math.oc may model training practice privacy privacy concerns private private data public stat.ml training work