Sparsity in neural networks can improve their privacy

2023/04/01 by Antoine Gonon, Gonon, Antoine, Léon Zheng +9
Computer Science · #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Stochastic Gradient Optimization Techniques

paper · pdf · doi:10.48550/arxiv.2304.10553

openalex publication_date 2023/04/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

This article measures how sparsity can make neural networks more robust to\nmembership inference attacks. The obtained empirical results show that sparsity\nimproves the privacy of the network, while preserving comparable performances\non the task at hand. This empirical study completes and extends existing\nliterature.\n

Related

• Can sparsity improve the privacy of neural networks?
• Membership Inference Attacks and Defenses in Neural Network Pruning
• Spurious Privacy Leakage in Neural Networks
• On the Privacy Risks of Spiking Neural Networks: A Membership Inference Analysis
• Better Membership Inference Privacy Measurement through Discrepancy