Steganalysis of content-adaptive JPEG steganography based on the weight allocation of filtered coefficients
Yi Zhang, Chunfang Yang, Xiangyang Luo, Fenlin Liu, Jicang Lu, Xiaofeng Song · 2017
Comparing with the steganalysis methods based on the feature sets assembled as histograms of filtered images, their improved versions incorporating the change probabilities of coefficients in the embedding domain provide more excellent performance for content-adaptive JPEG steganography, the weight allocation of feature statistical samples is the most important. In this paper, we propose a new weight allocation method in which the maximum change probability of corresponding correlation DCT coefficients is calculated as the weight of each filtered coefficient, and the final feature set is obtained by accumulating the weights in corresponding histogram statistical samples. Experimental results conducted on three modern content-adaptive JPEG steganographic schemes and the state-of-the-art steganalysis feature set indicate that the proposed method can improve the detection performance of original feature set markedly and is superior to the selection channel aware feature set.