基于图像频域统计矩特征的隐写分析方法  

Image Steganalysis based on Statistical Moments in Frequency Domain

作　　者：时书剑[1] 胡声丹[2] 毛家发[3] 

机构地区：[1]上海师范大学天华学院计算机系,上海201815 [2]北京邮电大学网络与交换技术国家重点实验室信息安全中心,北京100875 [3]上饶师范学院数学计算机系,江西上饶334001

出　　处：《中国电子商情（通信市场）》2010年第3期147-151,共5页

基　　金：由上海市教育发展基金会光计划项目（2008CGB21）资助.

摘　　要：针对互联网上大量的隐写软件利用图像隐藏秘密消息，本文提出了基于图像频域统计矩特征的隐写分析方法。首先，对图像进行三级小波分解，并对一级分解的高频子带再次分解，得到25个子带。然后，计算小波子带前3阶正向特征函数矩，并利用KPCA提取不重要核主元，最后用LibSVM进行分类。仿真结果表明该方法对F5和Jsteg隐写有较好的检测性能。For secret messages are hidden in the images by a large number of steganographic software on the Internet, a steganalysis method based on statistical moment in frequency domain is proposed in this paper. First, The image was decomposed by 3-level discrete wavelet transform, including the further decomposition of the first scale high frequency sub-bands. There were 25 sub-bands. Then, the top 3 positive characteristic function moments of wavelet sub-bands were calculated. And the least kernel principal components were extracted by using KPCA. Finally, LibSVM was adopted as the classifier. The simulation results show the proposed method performs well in attacking F5 and Jsteg.

关 键 词：信息安全 隐写分析 特征函数矩 KPCA 支撑向量机 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]