A combined deep learning training method based on a privacy protection technology

Abstract

The invention belongs to the technical field of artificial intelligence, and relates to a combined deep learning training method based on a privacy protection technology. The efficient combined deep learning training method based on the privacy protection technology is achieved. In the invention, each participant first trains a local model on a private data set to obtain a local gradient, then performs Laplace noise disturbance on the local gradient, encrypts the local gradient and sends the encrypted local gradient to a cloud server; The cloud server performs aggregation operation on all thereceived local gradients and the ciphertext parameters of the last round, and broadcasts the generated ciphertext parameters; And finally, the participant decrypts the received ciphertext parameters and updates the local model so as to carry out subsequent training. According to the method, a homomorphic encryption scheme and a differential privacy technology are combined, a safe and efficient deep learning training method is provided, the accuracy of a training model is guaranteed, and meanwhile a server is prevented from inferring model parameters, training data privacy and internal attacksto obtain private information.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and relates to a joint deep learning training method based on privacy protection technology. Background technique [0002] Federated Deep Learning allows parties to jointly train deep learning models without disclosing private datasets. Each participant trains the model independently on the private data set, and shares the training results such as gradients and parameters with other participants, thereby indirectly sharing their own training data. Compared with centralized deep learning, joint deep learning does not need to collect users' private data, which has higher efficiency and stronger security, and the participants hold the trained model for local prediction. [0003] Differential Privacy (Differential Privacy) is a cryptographic technique that removes individual features while retaining statistical features to protect user privacy. The Laplacian mechanism is often used t...

AI Technical Summary

Problems solved by technology

Although the use of secure multi-party computing ensures the privacy of users under large-scale users, the huge communication overhead limits the practical application

Typical public key encryption schemes (such as the Pallier homomorphic encryption scheme) present a huge computational overhead when a large amount of data is aggregated, resulting in a slow network training process

In addition, differential privacy technology protects data privacy by adding noise to training data or training gradients, in order to achieve security, resulting in a reduction in model accuracy

Images