Decentralized Federated Learning Based on Committees and Blockchain

TL;DR

This paper proposes a decentralized federated learning system utilizing blockchain and consensus algorithms to enhance privacy, security, and fairness in collaborative machine learning across communities.

Contribution

It introduces a novel decentralized framework for federated learning that incorporates blockchain technology and consensus mechanisms to address security and fairness issues.

Findings

System effectively supports decentralized federated learning

Federated learning maintains high model accuracy

Privacy protection mechanisms are validated

Abstract

Machine learning algorithms are undoubtedly one of the most popular algorithms in recent years, and neural networks have demonstrated unprecedented precision. In daily life, different communities may have different user characteristics, which also means that training a strong model requires the union of different communities, so the privacy issue needs to be solved urgently. Federated learning is a popular privacy solution, each community does not need to expose specific data, but only needs to upload sub-models to the coordination server to train more powerful models. However, federated learning also has some problems, such as the security and fairness of the coordination server. A proven solution to the problem is a decentralized implementation of federated learning. In this paper, we apply decentralized tools such as blockchain and consensus algorithms to design a support system that supports the decentralized operation of federated learning in an alliance environment, involving the exploration of incentives, security, fairness and other issues. Finally, we experimentally verify the performance of our system, the effect of federated learning, and the availability of privacy protection.