Steganalysis for Markov Feature of Difference Array in DCT Domain

In this paper, an effective steganalysis method based on Markov transition matrix is proposed to attacking the advanced JPEG steganographic schemes. As the horizontal and vertical scan modes failed to detect F5 steganography with low imbedding-rate, zigzag scanning pattern is developed to arrange both DCT (Discrete Cosine Transform) blocks and coefficients in each block. Difference computation was applied on the arranged JPEG 2-D array, which enhanced the variation caused by data embedding. Threshold and truncation techniques are utilized to make the computational complexity of the proposed method manageable. Bidirectional Markov matrix is exploited to capture the correlations between the adjacent coefficients in both intra-block and inter-block senses, which have been changed during data embedding. Features for steganalysis are derived from intra-block and inter-block Markov transition matrixes. The experiment works have demonstrated the effectiveness of the proposed scheme, which greatly improves the capability of steganalysis, the detection accuracy for each steganography at embedding-rate 0.05 exceeds 95%.