When Blockchain Meets Asynchronous Federated Learning

In the face of issues such as privacy leakage and malicious attacks, blockchain-based asynchronous federated learning emerges as a promising solution, not only capable of protecting user privacy and resisting malicious attacks but also outperforming its synchronous counterpart in terms of aggregation speed and robustness against low-performance devices. Our work focuses on systematically categorizing recent advancements in blockchain-based asynchronous federated learning. To delve deeper into the advantages of integrating blockchain with asynchronous federated learning, we first provide relevant introductions. Subsequently, we systematically classify the works based on the types of blockchain extensions and coupling approaches. Finally, we discuss the opportunities and challenges faced by blockchain-based asynchronous federated learning, aiming to elucidate future research directions.