Federation learning defense method and device, electronic equipment and storage medium

Abstract

The invention relates to the artificial intelligence technology, and discloses a federated learning defense method comprising the following steps: obtaining local model parameters from each participant of federated learning, constructing attack model parameters according to the local model parameters, obtaining a training data set according to the local model parameters and the attack model parameters, and obtaining a training data set according to the training data set and the attack model parameters; training a pre-constructed discriminator by utilizing the training data set to obtain an attack discriminator, judging a to-be-detected local model by utilizing the attack discriminator to obtain a good local model parameter, and performing aggregation operation on the good local model parameter to obtain a global model parameter; and sending the global model parameters to each participant. In addition, the invention also relates to a blockchain technology, and the global model parameters can be stored in nodes of a blockchain. The invention further provides a federated learning defense device, electronic equipment and a computer readable storage medium. According to the method, theproblem of poor defense effect for a specific attack mode can be solved.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a federated learning defense method, device, electronic equipment and computer-readable storage medium. Background technique [0002] Malicious attacks are a major challenge in the development of federated learning algorithms, and attacks from internal devices of federated learning systems are one of the main attack methods. Malicious devices can significantly reduce the performance of the global model by attacking the training data set or the model training process, thus making federated learning meaningless. In addition, some targeted attacks will have a huge impact on the performance of the global model in certain aspects . In different federated learning systems, the attack methods are also different. The existing defenses are mainly some artificially designed defense algorithms for data attacks. For attacks with specific attack methods, the defense e...

AI Technical Summary

Problems solved by technology

[0003] The present invention provides a federated learning defense method, device and computer-readable stor

Images