Blockchain-based Federated Learning Data Sharing Model

Federated learning has been widely researched and applied in various scenarios. In the context of mobile computing, federated learning protects users from exposing their private data while enabling collaborative training of global models for various practical applications. However, the security of federated learning has come under increasing scrutiny due to attacks from malicious clients or untrusted central servers on global models or user privacy data. To address these security concerns, this paper proposes a federated learning data sharing model based on blockchain and reputation values. This data sharing model utilizes IPFS for storing model parameters, with the model parameters stored in IPFS hash addresses to alleviate the storage pressure on the blockchain. Additionally, an access control mechanism is designed. A consensus mechanism based on reputation values is proposed, aiming to enhance model performance in the face of malicious attacks.