基于GAN密钥生成模型的数字图像加密算法优化  

Optimization of Digital Image Encryption Algorithm Based on GAN Key Generation Model

作　　者：夏群[1] 丁小峰[1] 夏珍 XIA Qun;DING Xiao-feng;XIA Zhen(Nanchang Institute of Science and Technology,Nanchang Jiangxi 330108,China;School of software,Nanchang University,Nanchang Jiangxi 330108,China)

机构地区：[1]南昌工学院,江西南昌330108 [2]南昌大学软件学院,江西南昌330108

出　　处：《计算机仿真》2025年第1期229-232,334,共5页Computer Simulation

基　　金：江西省教育厅科学技术研究项目(GJJ212509);江西省高校人文社会科学研究青年项目(JY20221)。

摘　　要：针对目前数字图像加密领域中存在的图像加密效果不佳、解密后图像清晰度降低等问题,提出基于GAN密钥生成模型的数字图像加密算法优化方法。通过细胞神经网络超混沌系统生成随机混沌序列,通过Wasserstien距离和惩罚梯度对生成对抗网络模型实行优化处理;将随机混沌序列输入至改进后的生成对抗网络中获取密钥,并结合实时动态置乱方法和随机相位傅里叶变换方法实现数字图像的加密处理。实验结果表明,所提算法的数字图像加解密效果更好、安全系数更高,是一种适用于实际应用的图像处理方法。Aiming at the problems of poor image encryption performance and reduced image clarity after decryption in the current field of digital image encryption,a digital image encryption algorithm optimization method based on GAN key generation model is proposed.Firstly,the hyperchaotic system in cellular neural network was used to generate random chaotic sequences.And then,Wasserstien distance and penalty gradient were used to optimize the generative adversarial network model.Moreover,the random chaotic sequences were input into an improved adversarial network to obtain a key.According to real-time dynamic scrambling method and random phase Fourier transform method,we achieved the encryption of digital images.The experimental results show that the proposed algorithm has better encryption and decryption effects and higher security for digital images,so it is suitable for practical application.

关 键 词：改进生成对抗网络 秘钥生成 随机混沌序列 随机相位傅里叶变换方法 图像加密算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]