Privacy-protecting multi-party deep learning computation proxy method under cloud environment

Abstract

The invention belongs to the technical field of cloud computing, and is to achieve data sharing under the premise of protecting privacy and deep learning application on the basis of data sharing. Thetechnical scheme adopted by the method is a privacy-protecting multi-party deep learning computation proxy method under a cloud environment. Each participant runs a deep learning algorithm based on arespective data set to compute a gradient parameter value, and uploads the gradient parameter that is encrypted by a multiplicative homomorphic ElGamal encryption scheme to a server; when uploading the gradient parameter to the cloud server, the participant simultaneously generates the signature of the parameter, and the signature meets polymerization, that is to say the cloud server can compute the gradient parameter and the signature; the cloud computing server computes the sum of gradient parameters of all users on a ciphertext, and returns the result back to the user, and the user acquiresthe final gradient parameter sum after decrypting, and verifies validity of the sum through checking whether the result and the polymerized signature are the effective message and signature. The method provided by the invention is mainly applied to cloud computing occasions.

Description

technical field [0001] The invention belongs to the technical field of cloud computing, and discloses a deep learning method for multi-party privacy protection in a cloud environment, and specifically relates to a multi-party deep learning computing proxy method for privacy protection in a cloud environment. Background technique [0002] Deep learning is a computational model composed of multiple processing layers for learning data representations with multiple levels of abstraction. The model starts from the raw data, and each layer can transform the representation of the previous layer into a representation of a more abstract level through a nonlinear transformation. Complex functions can be learned through the combination of enough such transformations. Recent advances in deep learning have significantly improved advanced techniques in the field of artificial intelligence, such as image recognition, speech recognition, Graves-year face detection, face recognition, and di...

AI Technical Summary

Problems solved by technology

[0008] To sum up, the existing technical problems are: the high computational complexity of the fully homomorphic encryption algorithm makes it difficult to widely apply multi-party deep learning based on the fully homomorphic encryption scheme

However, differential privacy also has the problem of privacy leakage, and the accuracy of the model is lower than that of conventional non-privacy-preserving multi-party deep learning models.

Images