Universal Image Steganalytic Method

In the paper we introduce a new universal steg- analytic method in JPEG file format that is detecting well- known and also newly developed steganographic methods. The steganalytic model is trained by MHF-DZ stegano- graphic algorithm previously designed by the same authors. The calibration technique with the Feature Based Steganalysis (FBS) was employed in order to identify sta- tistical changes caused by embedding a secret data into original image. The steganalyzer concept utilizes Support Vector Machine (SVM) classification for training a model that is later used by the same steganalyzer in order to identify between a clean (cover) and steganographic image. The aim of the paper was to analyze the variety in accuracy of detection results (ACR) while detecting testing steganographic algorithms as F5, Outguess, Model Based Steganography without deblocking, JP Hide&Seek which represent the generally used steganographic tools. The comparison of four feature vectors with different lengths FBS (22), FBS (66) FBS(274) and FBS(285) shows prom- ising results of the proposed universal steganalytic method comparing to binary methods.