ELITE: Defending Federated Learning against Byzantine Attacks based on Information Entropy

Yongkang Wang, Yuanqing Xia*, Yufeng Zhan

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

6 Citations (Scopus)

Abstract

Federated learning is a distributed machine learning paradigm, where physically distributed computing nodes collaboratively train a global model. In federated learning, workers usually do not share training data with others, and thus some workers are malicious workers, who can change parameters (e.g., weights/gradients) in their models to degrade the global model's training accuracy. This is generally called Byzantine attack. Existing solutions are either limited resistance to Byzantine attacks or not applicable to federated learning. In this paper, we propose ELITE, a robust parameter aggregation algorithm to defend federated learning from Byzantine attacks. Inspired by the observation that the parameters of malicious workers usually distract from the parameters of benign workers, we introduce entropy to efficiently detect malicious workers. We evaluate the performance of ELITE on image classification model training under three typical attacks, and experimental results show that ELITE can resist various Byzantine attacks and outperforms existing algorithms by improving the model accuracy at most up to 80%.

Original languageEnglish
Title of host publicationProceeding - 2021 China Automation Congress, CAC 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages6049-6054
Number of pages6
ISBN (Electronic)9781665426473
DOIs
Publication statusPublished - 2021
Event2021 China Automation Congress, CAC 2021 - Beijing, China
Duration: 22 Oct 202124 Oct 2021

Publication series

NameProceeding - 2021 China Automation Congress, CAC 2021

Conference

Conference2021 China Automation Congress, CAC 2021
Country/TerritoryChina
CityBeijing
Period22/10/2124/10/21

Keywords

  • Byzantine attacks
  • Federated learning
  • information entropy
  • robust

Fingerprint

Dive into the research topics of 'ELITE: Defending Federated Learning against Byzantine Attacks based on Information Entropy'. Together they form a unique fingerprint.

Cite this