Repository logo

Infoscience

  • English
  • French
Log In
Logo EPFL, École polytechnique fédérale de Lausanne

Infoscience

  • English
  • French
Log In
  1. Home
  2. Academic and Research Output
  3. Conferences, Workshops, Symposiums, and Seminars
  4. Towards Practical Homomorphic Aggregation in Byzantine-Resilient Distributed Learning
 
conference paper

Towards Practical Homomorphic Aggregation in Byzantine-Resilient Distributed Learning

Choffrut, Antoine
•
Guerraoui, Rachid  
•
Pinot, Rafaël
Show more
December 2, 2024
Middleware 2024 - Proceedings of the 25th ACM International Middleware Conference
25 ACM International Middleware Conference

The growing availability of distributed data has led to the increased use of machine learning (ML) algorithms in distributed topologies, where multiple nodes collaborate to train models under the coordination of a central server. However, distributed learning faces two significant challenges: the risk of Byzantine nodes corrupting the learning process by sending incorrect information, and the potential for a curious server to violate the privacy of individual nodes, even reconstructing their private data. While homomorphic encryption (HE) has been a promising solution for privacy preservation in distributed settings, its high computational cost, especially for high-dimensional ML models, has made it challenging to design robust (non-linear) Byzantine-resilient algorithms using HE. In this paper, we introduce SABLE, the first distributed learning protocol that is both Byzantine robust and fully homomorphic. SABLE utilizes the novel Homomorphic Trimmed Sum (HTS) operator, which efficiently implements the robust coordinate-wise trimmed mean, providing strong defenses against Byzantine nodes while ensuring data privacy through HE. Extensive experiments on standard ML tasks show that SABLE achieves practical execution times and offers ML accuracy comparable to non-private approaches, proving its effectiveness in real-world scenarios.

  • Details
  • Metrics
Type
conference paper
DOI
10.1145/3652892.3700783
Scopus ID

2-s2.0-85215525858

Author(s)
Choffrut, Antoine

Université Paris-Saclay

Guerraoui, Rachid  

École Polytechnique Fédérale de Lausanne

Pinot, Rafaël

Sorbonne Université

Sirdey, Renaud

Université Paris-Saclay

Stephan, John  

École Polytechnique Fédérale de Lausanne

Zuber, Martin

CryptoNext Security

Date Issued

2024-12-02

Publisher

Association for Computing Machinery, Inc

Published in
Middleware 2024 - Proceedings of the 25th ACM International Middleware Conference
ISBN of the book

9798400706233

Start page

431

End page

444

Subjects

Byzantine Robustness

•

Federated Learning

•

Homomorphic Encryption

•

Machine Learning

Editorial or Peer reviewed

REVIEWED

Written at

EPFL

EPFL units
DCL  
Event nameEvent acronymEvent placeEvent date
25 ACM International Middleware Conference

Hong Kong, Hong Kong

2024-12-02 - 2024-12-06

Available on Infoscience
February 10, 2025
Use this identifier to reference this record
https://infoscience.epfl.ch/handle/20.500.14299/246768
Logo EPFL, École polytechnique fédérale de Lausanne
  • Contact
  • infoscience@epfl.ch

  • Follow us on Facebook
  • Follow us on Instagram
  • Follow us on LinkedIn
  • Follow us on X
  • Follow us on Youtube
AccessibilityLegal noticePrivacy policyCookie settingsEnd User AgreementGet helpFeedback

Infoscience is a service managed and provided by the Library and IT Services of EPFL. © EPFL, tous droits réservés