The Ensemble Federated Learning System of Privacy Protection in Industrial Internet of Things Based on Blockchain
The data in industrial internet of things(IoT)devices contain valuable information.Federated learning technology can be used to train models and discover data information while protecting data privacy.However,practical applications of federated learning still face challenges such as node heterogeneity,data privacy leakage,and single point of failure on the server.Therefore,the privacy-preserving heterogeneous ensemble federated learning system based on blockchain is proposed.Firstly,the system uses ensemble learning to optimize heterogeneous federated learning,further improving the prediction accuracy.Seconelly,it introduces split learning,allowing nodes to jointly complete model training with mobile edge computing servers,and uses differential privacy technology to further protect data privacy.Finally,the trained model is stored on the blockchain,and the consensus algorithm of the blockchain further prevents attacks from malicious nodes,protecting the security of the model.Experiments show that the ensemble learning-optimized scheme can improve the prediction accuracy of heterogeneous federated learning and outperform traditional federated learning algorithms in heterogeneous node scenarios.
blockchainIndustrial Internet of Thingsdifferential privacyfederated learningsplit learningensemble learning
万方数据
维普
