TEE federated learning privacy protection method based on proxy re-encryption

The invention discloses a TEE federated learning privacy protection method based on proxy re-encryption, and belongs to the field of public key cryptosystem. Comprising the following steps: (1) establishing proxy re-encryption communication; (2) the server issues a federation strategy; (3) the participant performs local learning to obtain gradient parameters, encrypts the gradient parameters and sends the gradient parameters to the agent; (4) the agent side carries out re-encryption operation on the message and sends the message to the server side; (5) the server side decrypts the message; (6) the server side aggregates the gradient parameters, encrypts the gradient parameters and sends the gradient parameters to the agent side; (7) the agent performs re-encryption operation on the message and sends the message to the participant; (8) the participant decrypts the message to obtain a gradient parameter and continues to carry out model training; and (9) repeating the steps (3)-(8) until federal learning converg