Research On Image Universal Steganalysis Technology

With the development of steganography,the way of information hiding and the type of data become more and more diversified,the concealment and the security also become more and more higher.The digital image is one of the most commonly used as the carrier,and there are many kinds of steganography based on it.For this situation,it is one of the most difficulties that design a universal steganalysis algorithm to detect steganography based on digital image in this field.The main contribution of this dissertation:1,The paper proposed universal steganalysis features on based on image texture,the correlation of neighborpixel and coefficient,and the improvement of subtractive pixel adjacency matrix.The mostly spatial steganograhpy break the correlation of neighbor pixel and coefficient in some point,and will make the difference statistic features of image changed.The original subtractive pixel adjacency matrix model uses markov probability transfer matrix to calculate features.However,it is more better that use joint probability distribution matrix than markov probability transfer matrix to calculate the change of neighbor pixel and coefficient.The paper extracts some merge features from image based on difference matrix and improvement of subtractive pixel adjacency matrix for spatial steganalysis.The result of simulation and experiment shows that the merge features could detect many kinds spatial steganography.2,The paper proposed a transform domain steganalysis algorithm based on the improvement of Chen486.The original Chen486 steganalysis algorithm only extract the features of DCT sub-block in face of DCT sub-block.Based on the research and analysis of the original Chen486 model,this paper extracts the first-order statistical features of the DCT domain coefficients and DCT sub-block,except those,this paper also calculate the intra-block features of the DCT sub-blocks to form the merge features.The result of simulation and experiment shows that,the proposed algorithm has a better detection accuracy than the original Chen486.3,The paper proposed a universal steganalysis algorithm for color image.The paper applies above algorithms to the detection of steganography which based on color image.For detection color spatial domain steganography,the paper uses three ways to merge the features that extracted from different channel of image.For detecte color transform domain steganography,the paper uses two ways to merge the features that extract from different channel of image.The result of simulation and experiment shows that the universal steganalysis based on gray-scale image also could be apply to color image by using appropriate.