Statistical Modeling for LSB-Based Image Steganalysis: A Systematic Perspective

Steganalysis is a science of detecting possible hidden messages in an apparently innocuous cover medium, which is usually modeled as looking for a characteristic function that can discriminate effectively the stego from the cover. In this paper, we investigated the statistical models available for typical tools of LSB-based image steganalysis, focusing on their relations with the statistics of cover images, of secret messages, of operations acted upon the cover image, in an effort to theoretically form a systematic perspective which helps cast better insights into different steganalytic methods of inherent consistence. We proved the equivalence of some statistical models as results and compared the difference of effectiveness among these consistent statistical models.