SAFEFL: MPC-friendly Framework for Private and Robust Federated Learning

Till Gehlhar; Felix Marx; Thomas Schneider; Ajith Suresh; Tobias Wehrle; Hossein Yalame

Paper 2023/555

SAFEFL: MPC-friendly Framework for Private and Robust Federated Learning

Till Gehlhar

, Technical University of Darmstadt

Felix Marx

, Technical University of Darmstadt

Thomas Schneider

, Technical University of Darmstadt

Ajith Suresh

, Technical University of Darmstadt

Tobias Wehrle

, Technical University of Darmstadt

Hossein Yalame

, Technical University of Darmstadt

Abstract

Federated learning (FL) has gained widespread popularity in a variety of industries due to its ability to locally train models on devices while preserving privacy. However, FL systems are susceptible to i) privacy inference attacks and ii) poisoning attacks, which can compromise the system by corrupt actors. Despite a significant amount of work being done to tackle these attacks individually, the combination of these two attacks has received limited attention in the research community. To address this gap, we introduce SAFEFL, a secure multiparty computation (MPC)-based framework designed to assess the efficacy of FL techniques in addressing both privacy inference and poisoning attacks. The heart of the SAFEFL framework is a communicator interface that enables PyTorch-based implementations to utilize the well established MP-SPDZ framework, which implements various MPC protocols. The goal of SAFEFL is to facilitate the development of more efficient FL systems that can effectively address privacy inference and poisoning attacks.

Note: This is the full version of our research paper that has been accepted for publication at the 6th Deep Learning Security and Privacy workshop co-located with IEEE Security & Privacy (IEEE S&P) conference.

Metadata

Available format(s): PDF
Category: Implementation
Publication info: Published elsewhere. DEEP LEARNING SECURITY AND PRIVACY WORKSHOP 2023
Keywords: Federated Learning MPC Privacy
Contact author(s): till gehlhar @ stud tu-darmstadt de
felix marx @ stud tu-darmstadt de
schneider @ encrypto cs tu-darmstadt de
suresh @ encrypto cs tu-darmstadt de
tobias wehrle @ stud tu-darmstadt de
yalame @ encrypto cs tu-darmstadt de
History: 2023-04-24: approved; 2023-04-19: received; See all versions
Short URL: https://ia.cr/2023/555
License: CC BY

BibTeX

@misc{cryptoeprint:2023/555,
      author = {Till Gehlhar and Felix Marx and Thomas Schneider and Ajith Suresh and Tobias Wehrle and Hossein Yalame},
      title = {{SAFEFL}: {MPC}-friendly Framework for Private and Robust Federated Learning},
      howpublished = {Cryptology {ePrint} Archive, Paper 2023/555},
      year = {2023},
      url = {https://eprint.iacr.org/2023/555}
}