Decentralized federated learning based on blockchain: concepts, framework, and challenges

Abstract

Decentralized federated learning integrates advanced technologies, including distributed computing and secure encryption methodologies, to facilitate a robust and efficient mechanism for safeguarding data privacy and security during collaborative model training endeavors. The incorporation of blockchain technology into Federated Learning provides a transformative framework characterized by its inherent decentralization and data immutability, making it a focal point of contemporary research inquiry. The literature on the integration of blockchain technology with federated learning frameworks is presently deficient in comprehensive summary works. Such summaries are essential for advancing understanding of the implementation challenges and for guiding future research efforts in this domain. Therefore, in this work, we first summarize a typical decentralized federated learning framework based on blockchain and describe its operational workflow and its applications in the fields of the Internet of Things, the Internet of Vehicles, etc. A systematic summary of the challenges confronting this framework and an analysis of the solutions proposed to address these challenges are provided. Finally, this work provides insight into the possible future research directions.