Security calculation method based on joint learning

Abstract

The invention discloses a safety calculation method based on joint learning, and the method comprises the steps: building a joint learning model, importing different data of different objects into the joint learning model, carrying out the data feature extraction through a feature extraction network, and storing, two similar models are trained through an encryption algorithm according to two independent learning features which are different in object but are associated with each other, one model learns the other model, and value sharing is achieved; besides, the safety calculation method for joint learning is carried out based on an encryption state, when the two objects mutually reference the models of the opposite side for learning, privacy operation is carried out, data information of the two objects is protected, meanwhile, a confidential channel is provided for data joint learning of the two objects, the safety value is high, and the safety calculation method is suitable for popularization and application. The method is worthy of wide-range popularization and application.

Description

technical field [0001] The invention relates to the technical field of joint learning, in particular to a secure computing method based on joint learning. Background technique [0002] A typical example of joint learning is: two things repeating very close in time, and finally forming a connection in the brain, such as classical conditioning and operant conditioning belong to this type of learning. Existing applications of federated learning refer to solving machine learning problems through a series of algorithms. Their important details are different, but the basic idea is the same: although the server cannot collect data, it can collect the parameters of the model, the server coordinates the edge devices to participate in training, and each edge device has training data. Each edge device uses its own data to train a local model, and uploads its own parameters to the server encrypted or unencrypted, and the server broadcasts the collected parameters to each edge device by...

AI Technical Summary

Problems solved by technology

This method is still based on the learning process in the machine state, and cannot fully demonstrate the necessity of joint learning

[0003] Through the search of existing technologies, no benefits brought about by the practical application of joint learning have been found. Although simple joint learning is easy to be understood by the public, deep joint learning often has data leakage and data theft due to irregular technologies. Consequences of non-standard learning

Images