Steganalysis based on steganography pattern discovery

The goal of steganalysis algorithms is detection of stego images from clean images. Each steganography method based on its embedding mechanism puts a special pattern on the stego images. Finding this pattern in the images leads us to employ a classifier to be constructed specially for detecting stego images which are the results of a special steganography algorithm. In this paper, to have high detection accuracy, we propose an approach for Steganography Pattern Discovery (SPD). Our proposed approach employs an evolutionary method to extract the signature of stego images against clean images via fuzzy if-then rules. Based on the discovered knowledge, suitable trained models for steganalysis can be employed and stego images will be detected with high accuracy. Using SPD, we can predict the type of steganography method from a stego image. Employing SPD can enhance the approaches, which assume that a special steganography method is used. The effect of SPD before applying steganalysis methods has been investigated by some steganography and steganalysis techniques and it has been validated using some image databases. The results indicate that the pattern of a steganography method is extracted well and the type of steganography method used to make a stego image can be predicted with high accuracy.