基于复合混沌系统与人工神经网络学习的图像加密算法  

Image dynamic encryption algorithm based on one dimensional hybrid chaotic system and artificial neural network learning

作　　者：马凌 侯小毛 张福泉[2] 龚芝 

机构地区：[1]湖南信息学院电子信息学院 [2]北京理工大学软件学院

出　　处：《电子测量与仪器学报》2018年第8期109-116,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61373142);湖南省教育厅研究项目(XJK17CGD027)资助

摘　　要：为了提高密文的安全性,有效抗击明文攻击,提出了基于复合混沌系统与人工神经网络学习的图像动态加密算法。首先,基于非线性组合理论,利用Logistic、Tent与Sine映射来设计复合混沌系统,基于明文像素值,生成其初值,以输出混沌序列;将混沌序列作为输入层,引入人工神经网络,对其进行训练学习,消除其混沌周期性,输出神经网络序列,并定义集合混淆方法,对明文完成置乱;构造量化方法,对神经网络序列进行量化,获取密钥流;通过将混淆图像完成分类,联合密钥流,通过设计分段异扩散技术,改变其像素值,输出密文。实验结果显示,与当前图像加密机制相比,所提算法的输出密文具有更高的安全性与抗明文攻击能力。In order to improve the security of cipher to effectively resist plaintext attack,an image dynamic encryption algorithm based on one dimensional hybrid chaotic system and artificial neural network learning was proposed in this paper. First,the compound chaotic system was constructed by using the nonlinear combination mechanism to combine the Logistic,Tent and Sine map,and the chaotic sequence was outputted by using the pixel value to generate the initial value of compound chaotic system. Second,the chaotic sequence was viewed as input layer,then the neural network sequence was obtained by introducing the artificial neural network for learning to eliminate its chaotic periodicity,and the permutation image was got with the help of defining set confusion method. The quantitative method is used to quantify the neural network to obtain the key stream. Finally,the different sub diffusion technology was designed by sorting the confusion image and jointing in key stream to change the pixel value for outputting cipher. The experimental results show that this algorithm has higher security and stronger ability to resist plaintext attack compared with the current image encryption mechanism.

关 键 词：图像加密 复合混沌系统 人工神经网络 神经网络序列 集合混淆 分段异扩散 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构] TP391.41[自动化与计算机技术—计算机科学与技术]