Abstract: A new universal steganography detection algorithm based on Markov Chain（MC） theory is brought forward in this paper.This algorithm constructs Markov chain according to image neighborhood dependency，and captures diagonal elements from its empirical transition matrix as feature vector，and constructs a new discrimination function to detect whether the digital image is stego（hidden message） or not.Aiming at steganography based on global LSB，global DCT and global DWT，authors make detection experiment on the Matlab7.0 platform.According to experimental results authors improve on the algorithm further and make effect of detection better.Experimental results prove that the algorithm has better performances than the generic algorithms.

Key words: neighborhood dependency, Markov Chain（MC）, empirical transition matrix, Gaussian distribution, feature vector

摘要： 基于马尔可夫链（Markov Chain，MC）理论，提出了一种新的通用隐写检测算法。根据图像邻域相关的性质构造马尔可夫链，提取其经验转移矩阵的对角线元素作为特征向量，构造了一个新的判决函数作为检测秘密信息是否存在的依据。基于Matlab7.0平台，对全局LSB、DCT和DWT的隐写进行了检测实验。根据实验结果对算法进行了改进，使检测效果更优。结果证明：该算法的综合性能优于普通的检测算法。

关键词: 邻域相关性, 马尔可夫链, 经验转移矩阵, 高斯分布, 特征向量