Encryption protection system and method for neural network model relating to iteration and random encryption

Abstract

The invention belongs to the field of artificial neural network protection mechanism, and particularly relates to an encryption protection system and method for a neural network model relating to iterative and random encryption. The system comprises a data input module, an encryption module and an encryption data input module, an artificial neural network model module and a data output module; theencryption module comprises a structure conversion module and an iterative processing module; the iterative processing module comprises a password generation module, a password embedding module and asingle-layer convolution neural network model module; the password generation module comprises a fixed matrix generation module and a random matrix generation module. According to the invention, theprotective password can be embedded in the artificial neural network model under the condition that the calculation amount is not significantly increased and the performance of the artificial neural network is maintained, so that after the artificial neural network model is released, any copying, secondary development or modification cannot influence the protective password, and destroying the protective password can lead to a reduction in the performance of the artificial neural network model or may cause an effective output.

Description

technical field

[0001] The invention belongs to the field of artificial neural network protection mechanism, in particular to a neural network model encryption protection system and method involving iterative and random encryption. Background technique

[0002] Deep learning is the main technical solution for current artificial intelligence applications. The artificial neural network model trained with deep learning technology is the result of the labor of the original developers. However, in the process of publishing and applying the artificial neural network model, its network structure and node weights will be completely exposed to the outside world. After the artificial neural network model is released and / or applied by a third party, it is easy to be copied, re-developed or modified, resulting in damage to the rights and interests of the original developer. The protection schemes applicable to artificial neural network models in the prior art mainly include overall ne...

Images