A review on deep learning based image steganalysis

Deep learning is applied to image steganalysis and makes great progress. This paper presents a review on image steganalysis based on deep learning. According to the steganography scheme, these steganalysis methods are mainly classified as spatial domain steganalysis and jpeg domain steganalysis. The former approaches try to improve the steganalysis performance through network model design, fusion and learning strategy, while the latter methods attempt to perform better by designing appropriate network model or incorporating prior knowledge into the Convolutional Neural Network architecture. However, there are still several issues that need to be solved, on basis of which the future research direction is discussed.