针对序贯JSteg的提取攻击  

New Extracting Method for Sequential JSteg

作　　者：周涵[1] 刘九芬[1] 张卫明[1] 

机构地区：[1]信息工程大学信息工程学院,河南郑州450002

出　　处：《信息工程大学学报》2007年第3期265-268,共4页Journal of Information Engineering University

基　　金：国家自然科学基金资助项目(60473022);河南省自然科学基金资助项目(0511011300)

摘　　要：JSteg是一种基于JPEG图像的常见隐写算法。对序贯JSteg算法的提取攻击本质上是对连续嵌入的隐蔽信息起止点的估计问题。文章提出一种针对序贯JSteg的提取攻击方法,该方法首先通过多次实验选择合适的窗口,在窗口内执行2χ检验;然后通过滑动窗口,估计出起止点可能的存在区间;最后使用CPA(Change-Point Analysis)法,进一步精确估计隐蔽信息的起止点。实验结果表明,误差范围能达到真实起止点的-190bits^190bits。对序贯JSteg算法来说,该方法还可以作为一种隐蔽信息存在性的检验手段。JSteg is a popular steganographic algorithm based on JPEG. The aim of the extraction to sequential JSteg is to get the beginning and end of the secret message. This paper presents a new extracting method for sequential JSteg. First, we do many experiments to get proper number of the samples for chi-square test, and call this number for the slipping window. Second, we slip the window throughout the whole image. In every window, chi-square test must be carried out one time, and then the area which may conclude the beginning and end could be estimated. Third, we make sure the exact beginning and end through CPA （ Change-Point Analysis）. The results of experiments show that the error range can reduce to - 190bits - 190bits. For sequential JSteg, this method is also a detecting method of the secret message.

关 键 词：X^2检验 CPA JSteg 提取攻击 

分 类 号：TN918.1[电子电信—通信与信息系统] O157.4[电子电信—信息与通信工程]