Privacy Leakage of Adversarial Training Models in Federated Learning Systems

Jingyang Zhang; Yiran Chen; Hai Li

Privacy Leakage of Adversarial Training Models in Federated Learning Systems

Machine Learning 2022-02-23 v1 Cryptography and Security Computer Vision and Pattern Recognition

Authors: Jingyang Zhang , Yiran Chen , Hai Li

Abstract

Adversarial Training (AT) is crucial for obtaining deep neural networks that are robust to adversarial attacks, yet recent works found that it could also make models more vulnerable to privacy attacks. In this work, we further reveal this unsettling property of AT by designing a novel privacy attack that is practically applicable to the privacy-sensitive Federated Learning (FL) systems. Using our method, the attacker can exploit AT models in the FL system to accurately reconstruct users' private training images even when the training batch size is large. Code is available at https://github.com/zjysteven/PrivayAttack_AT_FL.

Keywords

privacy leakage adversarial training federated learning privacy attack deep neural networks image reconstruction

Cite

@article{arxiv.2202.10546,
  title  = {Privacy Leakage of Adversarial Training Models in Federated Learning Systems},
  author = {Jingyang Zhang and Yiran Chen and Hai Li},
  journal= {arXiv preprint arXiv:2202.10546},
  year   = {2022}
}

Comments

6 pages, 6 figures. Submitted to CVPR'22 workshop "The Art of Robustness"

Privacy Leakage of Adversarial Training Models in Federated Learning Systems

Abstract

Keywords

Cite

Comments

Related papers