Federal learning method and system based on multiple keys

The invention provides a federated learning method and system based on multiple keys. The method comprises the steps that a cloud server receives model gradient ciphertexts sent by all participants; calculating the model gradient ciphertext by using the server proxy key to generate a re-encrypted ciphertext; aggregating all re-encrypted ciphertexts to obtain an aggregated ciphertext group; decrypting the aggregated ciphertext group by using a server private key to obtain aggregated gradient data, and encrypting the aggregated gradient data to obtain an aggregated gradient ciphertext; and distributing the aggregation gradient ciphertext to each participant. By using the scheme of the invention, the security of privacy data of each participant in federal learning can be improved. 本发明提供一种基于多密钥的联邦学习方法及系统，该方法包括：云服务器接收各参与方发送的模型梯度密文；利用服务器代理密钥对模型梯度密文进行计算，生成重加密密文；聚合所有重加密密文，得到聚合密文组；利用服务器私钥对聚合密文组解密得到聚合梯度数据，对聚合梯度数据进行加密，得到聚合梯度密文；将聚合梯度密文分发给各参与方。利用本发明方案，可以提升联邦学习中各参与方隐私数据的安全性。