New YASS Steganalysis Method Based On JPEG Images

Because of JPEG as one of the most popular image formats on the Internet, the information hiding technology using JPEG images as cover media has developed rapidly. We will mainly discusses the JPEG steganalysis technology in this paper.YASS is a newly developed JPEG steganographic method. Through embedding data in the 8×8 blocks which do not coincide with the 8×8 grid used in JPEG compression, it effectively resists attacks of today's JPEG steganalyzers.Given the fact that YASS algorithms need artificial matching classifier at present, an analysis method aiming at the size of YASS big block is proposed in this paper, based on the phenomenon that QIM will make the coefficients distribution change and make the coefficients value of zero increase. The zero rounded re-quantized DCT coefficients features of stego images are extracted to automatic match classifiers. This algorithm can be widely applied to the existing YASS steganalysis methods. Experimental results show that, we can accurately analyze the big block size of stego images using support vector machine.In this paper, we also describe a new steganalysis method for attacking YASS, which modifies the difference-calibrated 274-dimensional JPEG feature Markov process feature vector with Cartesian calibration. To illustrate the validity of this method, image libraries with different big block sizes, quality factors and embedding capacity are separately carried on in the experiments. We demonstrate experimentally F5, Outguess, MB1 and YASS algorithms also can be reliably detected by this method even for small embedding rates.