An asynchronous federated learning-assisted data sharing method for medical blockchain

Currently, medical blockchain data sharing methods that rely on federated learning face challenges, including node disconnection, vulnerability to poisoning attacks, and insufficient consideration of conflicts of interest among participants. To address these issues, we propose a novel method for data sharing in medical blockchain systems based on asynchronous federated learning. First, we develop an aggregation algorithm designed specifically for asynchronous federated learning to tackle the problem of node disconnection. Next, we introduce a Proof of Reputation (PoR) consensus algorithm and establish a consensus committee to mitigate the risk of poisoning attacks. Furthermore, we integrate a tripartite evolutionary game model to examine conflicts of interest among publishing nodes, committee nodes, and participating nodes. This framework enables all parties involved to make strategic decisions that promote sustainable data-sharing practices. Finally, we conduct a security analysis to validate the theoretical effectiveness of the proposed method. Experimental evaluations using real medical datasets demonstrate that our method outperforms existing approaches.