Composite privacy protection method and system, computer equipment and storage medium

Abstract

The invention provides a composite privacy protection method and system, computer equipment and a storage medium; the method comprises the steps: generating an encryption public key of a client and a decryption private key of a server through a trusted third party in advance according to the weight of the client; sending, by the server, a first model and a first model parameter to the client; training the first model by the client according to local data, and updating the first model parameter to a second model parameter; performing a differential privacy algorithm to add the noise to the second model parameter by the client to obtain a third model parameter; according to the encryption public key, performing, by the client, function encryption on the third model parameter to obtain an encryption model and sends the encryption model to the server; and according to the decryption private key, decrpyting, by the server, the encryption model to obtain a global model. According to the method, privacy protection is enhanced, and model service quality and learning efficiency are improved.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a compound privacy protection method, system, computer equipment and storage medium oriented to federated learning. Background technique [0002] With the rapid development of artificial intelligence, the use of machine deep learning models for training and classification prediction is widely used. However, due to the limited data of each user, the accuracy of machine learning is reduced, and for the protection of data privacy and security, each The existence of problems such as the inability to directly exchange data between users limits the development of machine learning. Subsequently, a federated learning that allows full use of decentralized training equipment for model training without the need for original data came into being. . Although federated learning provides convenience for further breaking data islands and providing more accurate services,...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a compound privacy protection method oriented to federated learning, which overcomes the problem that the existing privacy protection method ignores the protection of the weight of the client, and achieves the effect of protecting the privacy of the client data source and preventing the threat of data-level privacy leakage At the same time, it overcomes the problem that the existing technology ignores the model service quality and learning efficiency of federated learning, and achieves the effect of preventing the threat of content-level privacy leakage and ensuring the model service quality and learning efficiency.

Images