JPEG Images Steganalysis Research Based On Bayes Decision

With the rapid development of information technology and web communication technology, steganalysis is becoming an important issue in the field of information hiding. As the antithesis of steganography, steganalysis aims at detecting the existence of secret information, estimating the length of the secret information and even extracting it. Universal steganalysis is implemented when the steganographic method is unknown and only the detected object is known, which is the inevitable trend of the steganalysis development. JPEG image is the most common image storage format, and more and more steganography systems select it as carrier image. Therefore, it is of great practical significance to research JPEG image steganalysis.In theory, detection rate will be raise if features are extracted from several angles for JPEG image steganalysis. This paper proposes a multi-feature JPEG image steganalysis algorithm, which includes discrete cosine transform(DCT) feature, Markov feature and multi-direction probability transition matrix feature. They are extracted from three angles based on DCT coefficients and are all sensitive to information embedding. The results of steganalysis experiments show that the proposed multi-feature algorithm is more outstanding compared with other steganalysis algorithms.After further analysis to the multi-feature algorithm, it could be found that the multi-feature algorithm does not completely play the role from experimental data because the feature dimension is too high. To solve this problem, this paper proposes a new universal steganalysis model for high division features. It divides multi-feature algorithm into five subspaces. From the test results of each subspace, the final result is obtained by the minimum risk Bayes decision fusion rule. Experiment results show that this model could effectively avoids the high dimension feature problems, and the detection rate to steganography algorithms increases apparently.