2022/05/25 by Yunhao Yang, Yang, Yunhao, Parham Gohari +3
Computer Science · #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Privacy-Preserving Technologies in Data
paper · pdf · doi:10.48550/arxiv.2205.12430
openalex publication_date 2022/05/25 · openalex created_date 2022/06/13 · openalex updated_date 2026/07/28
We study the privacy risks that are associated with training a neural network's weights with self-supervised learning algorithms. Through empirical evidence, we show that the fine-tuning stage, in which the network weights are updated with an informative and often private dataset, is vulnerable to privacy attacks. To address the vulnerabilities, we design a post-training privacy-protection algorithm that adds noise to the fine-tuned weights and propose a novel differential privacy mechanism that samples noise from the logistic distribution. Compared to the two conventional additive noise mechanisms, namely the Laplace and the Gaussian mechanisms, the proposed mechanism uses a bell-shaped distribution that resembles the distribution of the Gaussian mechanism, and it satisfies pure ε-differential privacy similar to the Laplace mechanism. We apply membership inference attacks on both unprotected and protected models to quantify the trade-off between the models' privacy and performance. We show that the proposed protection algorithm can effectively reduce the attack accuracy to roughly 50%-equivalent to random guessing-while maintaining a performance loss below 5%.