Using Higher Order DCT Difference to Effective Improve Markov Process Based JPEG Steganalysis Detection Rate

In recent, a Markov Process based approach has been proposed to effective attacking JPEG Steganography, and the experimental works had demonstrated that scheme has outperformed the existing steganalyzers in attacking JPEG steganography. To improve that original Markov Process based approach, several variants based that scheme has been proposed and experimented including using JPEG 1D array, using higher order difference of JPEG DCT coefficients, using inter-block 1D array and separating the DCT coefficients to several groups. Support vector machines (SVM) are utilized as classifier. The experimental works are presented to demonstrate that using intra-block JPEG 2-D array and second order difference can significantly improve detection rate.