Universal Steganalysis Using Feature Selection Strategy for Higher Order Image Statistics
The purpose of image steganalysis is to detect the presence of hidden message in cover photographic images. Supervised learning is an effective and commonly used method to cope with difficulties of unknown image statistics and unknown steganography. Present paper proposes; a universal approach for steganalysis for detecting presence of hidden messages embedded within digital images. This paper describes wavelet like decomposition to build higher order statistical model of natural images. Feature selection technique like ANOVA is used to select relevant features. SVM are then used to discriminate between clean and stego images. Study of the effect of relevant features on classification accuracy may help to improve the complexity.