Privacy-enhanced federated learning framework for power data analysis
To address the security threat of model update attacks on the deployment and application of federated learning in smart grids,this article proposes a secure and efficient federated learning framework for power data analysis based on cloud edge collaboration framework and privacy computing technology.By adding noise to the training model parameters of the client through differential privacy mechanism,the security of the model parameters during the training process is protected;Using secret sharing algorithm to securely aggregate noise model parameters,while ensuring fast convergence of the model,to protect power data and local model parameters.Theoretical analysis and experimental results indicate that this method can significantly improve the privacy of power data and shared model parameters.
NETL
NSTL
万方数据
