关键词: 联邦学习；区块链；差分隐私；激励机制；异常检测 

Abstract: A random verification blockchain construction and privacy protection method for federated learning was proposed to address the issues of local device model gradient leakage, the ability of centralized server devices to exit at will, and the inability of global models to resist malicious user attacks in existing federated learning models. Firstly, randomly elect blockchain leadership nodes were introduced by verifiable hash functions, ensuring the fairness of node block output; Secondly, a verification node cross detection mechanism was designed to defend malicious node attacks; Finally, based on differential privacy technology, blockchain nodes were trained, and incentive rules were constructed based on the contribution of nodes to the model to enhance the training accuracy of the federated learning model. Experimental results showe that compared to existing federated learning methods such as Google FL and LDP-FL, the proposed method can ensure good accuracy while improving the security of the model. The proposed mehod achieves 80% accuracy for malicious node poisoning attacks, which is 61 percentage points higher than Google FL, and the gradient matching loss is 14 percentage points higher than Google FL. It can be seen that the proposed method has better security and robustness.

Key words: federated learning, blockchain, differential privacy, incentive mechanism, anomaly detection