Research of multi-direction transition probability matrices algorithm for JPEG steganalysis
Jun Xiao, Bin Qin, Yusheng Sun, Xiao Xue Xu · 2012
With the rapid development of information technology and Web communication technology, steganalysis is becoming an important issue in the field of information hiding. A multi direction transition probability Matrices algorithm based on Bayes decision is proposed for JPEG images steganalysis. First, the multi-direction transition probability matrices are presented for image steganalysis. It captures the correlation of DCT coefficients from eight directions. 450-dimensional feature is extracted from intra-blocks and inter-blocks, and the feature catches the changes of the image when the secret information is embedded. Then, as the feature dimension is too high, a new universal steganalysis model and Bayes decision rule are adopted. Finally, the SVM is utilized as classifier. Experiment results demonstrate that multi-direction transition probability matrices can effectively detect Jsteg, F5, OutGuess and MB1 steganagraphies. And especially for F5 steganagraphy, the detection ratio increases apparently compared with other steganalysis methods. With Bayes decision, the proposed model could effectively avoid the high dimension feature problems.