Towards Blockchain-Based Fair and Trustworthy Federated Learning Systems

Ahmed Mukhtar Dirir, Khaled Salah, Davor Svetinovic

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

Abstract

Recently, Federated Learning (FL) gained considerable popularity as it offers an isolated and privacy-preserving mechanism to train Machine Learning models on unseen data. However, the use of the cloud server to build the global model might raise fairness and trust concerns since any FL server might try to regenerate the original data of some users. In this chapter, we review the key trust requirements for Decentralized Federated Learning (DFL) and provide the analysis in terms of fairness, trust, and privacy. We also present and compare the existing blockchain solutions for the development of fair and trustworthy FL systems.

Original languageBritish English
Title of host publicationStudies in Computational Intelligence
PublisherSpringer Science and Business Media Deutschland GmbH
Pages157-171
Number of pages15
DOIs
StatePublished - 2021

Publication series

NameStudies in Computational Intelligence
Volume965
ISSN (Print)1860-949X
ISSN (Electronic)1860-9503

Keywords

  • Blockchain
  • Decentralization
  • Fairness
  • Federated learning
  • Smart contract
  • Trust

Fingerprint

Dive into the research topics of 'Towards Blockchain-Based Fair and Trustworthy Federated Learning Systems'. Together they form a unique fingerprint.

Cite this