Byzantine robust federated learning-oriented user data privacy protection system and method

The invention provides a user data privacy protection system and method oriented to Byzantine robust federated learning, each server uses an RPCA method improved through a SecQR protocol and a SecEigen protocol to carry out security dimension reduction processing on encrypted updated local parameters to obtain a dimension reduction result, and uses a clustering algorithm to cluster the dimension reduction result to obtain a clustering result, and the clustering result is used for clustering the encrypted updated local parameters to obtain the user data privacy protection system and the user data privacy protection method oriented to the Byzantine robust federated learning. And the global model is collaboratively updated. According to the method, client privacy protection is emphasized, the SecQR protocol and the SecEigen protocol are applied to encrypted data processing, the security of sensitive information is guaranteed, good robustness is achieved to cope with potential threats, malicious model parameters