Universal image steganalytic method based on binary similarity measures

This paper is focused on comparison of two steganalytic methods that are able to detect embedding of secret message using popular and novel steganographic algorithms in JPEG images. First one is based on binary similarity measures and second method exploits 66 or 274 statistical features in transform domain from cover and stego images. The tested universal model was trained using novel steganographic method previously proposed by the same authors. Subsequently, the obtained statistical parameters from cover and stego images are used for training of model that is applied on detection of secret message in testing phase. The aim of this paper was to examine the accuracy detection (ACR) of these novel universal steganalytic models for popular steganographic tools in static images and also to compare time consumption in process of extraction parameters and training of model.