一种基于神经网络的图像加密算法  

An image encryption algorithm based on neural network

作　　者：黎学军 苗苗[1] 李庆宇 刘俊俏 LI Xuejun;MIAO Miao;LI Qingyu;LIU Junqiao(School of Information Science and Engineering, Dalian Polytechnic University, Dalian 116034, China;Shenyang Military Agency of the Army Headquarters in Dalian, Dalian 116033, China)

机构地区：[1]大连工业大学信息科学与工程学院,辽宁大连116034 [2]陆装沈阳军代局驻大连地区军代室,辽宁大连116033

出　　处：《大连工业大学学报》2021年第5期358-366,共9页Journal of Dalian Polytechnic University

基　　金：辽宁省教育厅科学研究项目(J2020114).

摘　　要：混沌加密为当今数字图像的机密性保护提供了一种实用的方法。针对图像信息在网络中的安全传输问题,设计了一种基于激光混沌系统与神经网络的图像加密算法。通过相图、李雅普诺夫指数谱和分岔图分析,证实了单模激光动力系统的混沌特性。通过BP神经网络对原始图像进行压缩,对压缩后的图像采用单模激光混沌系统生成的混沌序列进行像素置乱和扩散操作,得到密文图像,再通过直方图、相邻像素相关性、抗差分攻击、信息熵等方面分析所提算法的安全性。结果表明,所提出的图像加密算法不但能够节约密文传输时所用的信道带宽,而且具有良好的加密效果和较高的安全性能。Chaotic encryption provides a practical method for the confidential protection of digital images nowadays.To address the problem of the security transmission of image information in the network,an image encryption algorithm based on laser chaotic system and neural network was designed.The analysis of phase diagrams,Lyapunov exponential spectra and bifurcation diagrams confirmed the chaotic characteristics of the single-mode laser power system.The original images were firstly compressed by the BP neural network,and then the compressed images were processed with pixel scrambling and diffusion to obtain the ciphertext images,using the chaotic sequences generated by the single-mode laser chaotic system.The security of the proposed algorithm was analyzed by the histograms,correlation of adjacent pixels,anti-differential attack and information entropy.The experimental results showed that the proposed image encryption algorithm can not only reduce the channel band width in ciphertext transmission,but also possess good encryption function and high security performance.

关 键 词：激光混沌系统 神经网络 ARNOLD变换 图像加密 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]