Neural Network Model Encryption Protection System and Method Involving Specified Frequency Band Encryption

Abstract

The invention belongs to the field of an artificial neural network protection mechanism, and relates to a neural network model encryption protection system and method involving specified frequency band encryption. The neural network model encryption protection system comprises a data input module, an encryption module, an encrypted data input module, an artificial neural network module and a dataoutput module, wherein the encryption module comprises a frequency domain transformation module, a frequency domain data processing module and a frequency domain inverse transformation module; and thefrequency domain data processing module comprises a frequency domain specifying module, a password generating module and a password embedding module. The neural network model encryption protection system can embed a protective password into an artificial neural network model trained based on input data through the encryption for the input data under the condition of not significantly increasing the calculation amount and mainlining the performance of the artificial neural network. In addition, a method for input data encryption is realized through domain transformation, the hiding and anti-attacking performance of password information are improved, and the safety of the trained artificial neural network model is enabled to be further improved.

Description

technical field [0001] The invention belongs to the field of artificial neural network protection mechanisms, and in particular relates to a neural network model encryption protection system and method involving encryption of designated frequency bands. 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 labor result of the wisdom of the initial 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, redeveloped or modified, resulting in damage to the rights and interests of the original developers. The protection schemes suitable for artificial neural network models in the prior ar...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the above-mentioned defects of the existing artificial neural network protection mechanism, the present invention provides a neural network model encryption protection system and method involving specified frequency band encryption. The system and method are designed not to affect the structure and performance of the artificial neural network. Embedding a protective password in the raw input data used to train an artificial neural network such that the trained artificial neural network cannot process raw input data without a password embedded or properly handle raw input data embedded with an incorrect password

Images