一种针对BLOCK-DCT隐写的隐写分析改进算法

摘 要: 由于BLOCK-DCT隐写算法对平稳性质的原始图像引入了非平稳性，使得隐写图像块内部的相邻像素对与块之间的像素对的分布存在着差异。而在未嵌入消息的原始图像中它们的分布是一致的。由此利用Kolomogrov-smirnov test(k-s test)去判定图像是否隐藏信息.可是采用4邻域象素对的直接象素差的分布，在k-s test中并不怎么理想，两类概率分布之差的上界D都比较小。我们通过引入差的绝对值和8领域像素对来扩大2类象素对分布差异的方法来提高D的值，使得一些在以4邻域直接象素差分布方法中检测不出的隐写图像在改进的方法中得以检测。理论和实验都证明我们的想法是有效的。

An Improved Steganalysis Based on Block-DCT Steganography

BLOCK-DCT based information embedding methods introduce distinctive non-stationarites into the stego-image,which makes the difference distributions between neighboring pixel intensities in one block and across two blocks.While in the cover image these two distributions are indeed remarkably similar.So we use Kolomogrov-smirnov test on these two neighboring pixel populations to test binary hypothesis.In order to improve detective ability,we add absolute value and 8-domain to pairs of neigh- boring pixel.Theories and experiments both prove that our method is efficiency.