Image Steganalysis Based On Pattern Recognition

Steganography and steganalysis are important issues in the field of information security during the era of Internet. They are drawing more and more attentions from governments, military, security agents and research institutes all over the world. The primary task of Steganalysis is to make statistical analysis of the multimedia signals and determine whether secret information has been hidden in.In this thesis, after we analyze popular data hiding and/or steganography methods and the statistical variance of the cover image brought by data hiding process, we propose a new image steganalysis scheme which employs statistical pattern recognition methods and takes the statistical moments in the frequency domain of wavelet histogram of the image and the prediction-error image. And also, we prove—theoretically and experimentally—the effectiveness of the proposed features. The experiments towards data hiding methods in BMP images and JPEG steganography images show that the proposed steganalysis scheme is obviously superior to the prior arts. The research work of this paper points out a promising way to general image steganalysis.The main research work and contributions are as follows:1) Analyzed popular data hiding and steganographical methods, pointed out their influences on the statistics of the cover image so as to give directions for feature selection of steganalysis.2) Proposed, for the first time, that statistical moments in the frequency domain of wavelet histogram of the image be taken as features for steganaysis. Defined, for the first time, the statistical moment in the frequency domain of histogram. And Pointed out its general change while imbedding data.3) Proposed that using techniques of wavelet decomposition and prediction error, features be extracted from wavelet sub-bands and prediction-error images, and hence the relevant information of the neighboring pixels beextracted and a high dimension feature vector be formed so that the performance of steganalysis be enhance greatly.4) Proved—theoretically and experimentally—the effectiveness of the moments in frequency domain of histogram as features for steganalysis. Pointed out the moments in frequency domain of histogram are more effective than the moments in spacial domain of histogram and thus pointed out a could-be way to general image steganalysis.5) Proposed that features be extracted from spacial pixel values for BMP image and from DCT coefficients for JPEG images. The results of experiments towards BMP and JPEG image steganographical methods show that the proposed image steganalysis scheme is effective and superior.