Universal Image Steganalytic Method

Abstract

In the paper we introduce a new universal steganalytic method in JPEG file format that is detecting well-known and also newly developed steganographic methods. The steganalytic model is trained by MHF-DZ steganographic algorithm previously designed by the same authors. The calibration technique with the Feature Based Steganalysis (FBS) was employed in order to identify statistical 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 promising results of proposed universal steganalytic method comparing to binary methods.