Neural network-based white-box encryption method and system

By simulating the encryption and decryption operations of the AES algorithm through neural networks, a white-box encryption structure is constructed, which solves the problem of insufficient security and anti-attack capability of existing white-box encryption methods and realizes a white-box encryption scheme with high security and resistance to quantum computer attacks.

CN116760529BActive Publication Date: 2026-06-23ZHEJIANG LAB
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
ZHEJIANG LAB
Filing Date
2023-05-31
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing white-box encryption methods are at risk of key leakage in white-box environments and are difficult to resist various attack variations, resulting in insufficient security and resistance to attacks.

Method used

The encryption and decryption operations of the AES algorithm are simulated by using a neural network. By obfuscating and replacing the core steps of AES through a multi-layer neural network, a white-box encryption structure is constructed, including operations such as initial obfuscation, row shifting, column obfuscation and round key addition, forming an uninterpretable mapping relationship.

Benefits of technology

It improves the security and attack resistance of white-box encryption, achieves high security and reliability in a white-box environment, has the ability to resist cracking and analysis by quantum computers, and supports hardware acceleration.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116760529B_ABST
    Figure CN116760529B_ABST
Patent Text Reader

Abstract

The white-box encryption method based on neural network includes a confusion technique and a white-box AES encryption and decryption process, and the core steps of the AES scheme with a fixed key are summarized into multiple mapping tables, and the neural network is used to replace the encryption and decryption operation through overfitting training of the tables to achieve the confusion effect. The white-box AES encryption process is divided into two stages, the round key of AES each round is added with a byte substitution result, a random bijection is used for permutation to protect the key, and the combined mapping is confused by a neural network; secondly, the corresponding permutation recovery operation is introduced in the column confusion operation, and a randomly generated key byte is XORed after the finite field multiplication operation, and the above operation is replaced by a neural network. The decryption process is the same. The present application is highly combined with the neural network, and for the first time, the neural network is used to confuse part of the steps in the block encryption, and through the unexplainable nature of the neural network, a reliable black box environment is provided for part of the operation, and the present application has the advantages of good security, strong attack resistance and high expansibility.
Need to check novelty before this filing date? Find Prior Art