A New Steganalysis Method for Steganographic Images on DWT Domain

In this paper, we introduce a new method for steganalysis of grey-scale images. First, we analyzed the effect of various steganographic processes on the statistical properties of the image. So we extracted the optimal features from the images, which have high ability in make differentiated between two groups of normal and stego images. In this method, high order statistics in discrete wavelet transform (DWT) coefficients are used. Then the pre-processing of principal component analysis (PCA) is done on extracted features. The support vector machines (SVM) is used to classify image segments into stego or non-stego cases. The proposed method with comprehensive look into current steganographic techniques in DWT domain is able to detect the presence of hidden messages with more than 90% accuracy in different embedded rates. KeywordsSteganalysis, statistics features, principal component analysis (PCA), support vector machines (SVM).