Gauss-Markov载体下扩频隐写的抗检测性能分析  

Security of Spread-Spectrum Stegnaography for Gauss-Markov Covers

作　　者：孙怡峰[1,2] 刘粉林[1] 

机构地区：[1]信息工程大学信息工程学院,郑州450002 [2]信息工程大学电子技术学院,郑州450004

出　　处：《数据采集与处理》2010年第4期462-468,共7页Journal of Data Acquisition and Processing

基　　金：国家高技术研究发展计划("八六三")计划(2006AA01Z409)资助项目;河南省基础与前沿技术研究计划(082300410150)资助项目

摘　　要：为了研究载体数据相关性等对数字隐写安全性的影响,使用Gauss-Markov过程建模载体求解了扩频隐写系统的Kullback-Leibler divergence(KL散度)。首先,建立了以KL散度为度量标准的隐写系统安全性能分析框架;然后,选择Gauss-Markov过程建模载体,得到了扩频隐写系统的KL散度计算公式;分析相关系数、嵌入噪声相对强度、可用数据总数等参数对KL散度的影响,发现载体数据相关性越强,嵌入噪声强度越大,以及相同数据使用率下,载体可用数据总数越大,隐写系统抗检测性能越差。以图像为载体的实验结果验证了该方法的正确性。The effect of the cover data dependency on steganography security is investigated by evaluating the Kullback-Leibler （KL） divergence of spread-spectrum （SS） steganography with modeling cover as Gauss-Markov process. Firstly, a framework for evaluating the security of steganography is established. The framework uses KL divergence as the gauge of the security. Then, the KL divergence expression of SS steganography is obtained by modeling cover as Gauss-Markov process. The effect on security of the parameters, such as the correlation coefficient, the intensity of stego-noise, and the number of cover data, etc. , is investigated by nu- merically calculating KL divergence. Results show that the higher the cover data dependency is, the stronger the intensity of stego-noise is; with the same embedding rate, the larger the number of cover data is, the poorer the steganography security is. Experiments using digital image covers validate these viewpoints.

关 键 词：数字隐写 隐写分析 安全性 Gauss—Markov过程 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]