Weight encryption/decryption method and system aiming at neural network processing

Abstract

The invention relates to a weight encryption / decryption method and system aiming at neural network processing. The method comprises the following steps: encrypting trained original weight data through an encryption algorithm to generate the encrypted weight data; reading an original key and decrypting the encrypted weight data according to a decryption algorithm to generate the original weight data; reading an instruction needing to be executed, analyzing the instruction to obtain a storage address, and acquiring the input data from the storage address; receiving the input data and the original weight data, executing a neural network computing operation according to the instruction, and outputting a computing result. By use of the method and system disclosed by the invention, the weight data is encrypted through an off-chip encryption system, and the encrypted weight is decrypted by use of an on-chip decryption unit, thereby protecting the weight data while realizing the neural network computing.

Description

technical field [0001] The invention relates to the field of hardware neural network model calculation acceleration, in particular to a weight encryption and decryption method and system for neural network processing. Background technique [0002] Deep learning technology has developed rapidly in recent years, and it has been widely used in solving advanced abstract cognitive problems, such as image recognition, speech recognition, natural language understanding, weather prediction, gene expression, content recommendation and intelligent robots. And has excellent performance, so it has become a research hotspot in academia and industry. [0003] Deep neural network is one of the perception models with the highest level of development in the field of artificial intelligence. This type of network simulates the neural connection structure of the human brain by building a model, and describes the data features layered through multiple transformation stages, providing image, vide...

AI Technical Summary

Problems solved by technology

[0005] As a neural network computing hardware acceleration method, the neural network processor can perform calculations on the neural network model, but the existing neural network processor does not have an effective protection mechanism for the above-mentioned weight security issues, and there will be weight value leakage and flooding issues

Images