Research On Content-based Image Steganalysis

With the rapid development of computer science,multimedia information security has become the immediate concern of current research focus.As one of the important branches,digital image steganalysis is committed to detect the existence of hidden information,estimate its quantities,destroy or intercept it.Because of the complexity and diversity of natural scenes,the statistical characteristics of different content images are different from each other,and the degree of change of statistics before and after steganography is also different.Therefore,the performance of steganalysis is related not only to the embedding strategy of secret information,but also to the content of image itself.In view of the phenomenon,this dissertation focuses on the content-based image steganalysis,and the main contributions are summarized as follows.By analyzing the complexity of JPEG images,a novel approach based on pre-classification and feature extension is proposed.Firstly,according to the distribution of DCT coefficients,the training samples are divided into different categories so that the images contained in each sub-library have the same or similar statistical characteristics.Secondly,the deficiency of existing detection features is analyzed and effectively extended to improve the sensitivity of the feature to the statistical characteristics of the images before and after steganography.After that,the extended feature set is extracted from each category respectively to build a classifier.Given a testing image,its category is determined according to the minimum distance from its complexity to the average value of each category.Then,the feature set is extracted and sent to the corresponding classifier.Experimental results show that this method has better detection performance for steganographic algorithms with different embedding schemes.In addition,aiming at the shortcomings of traditional methods,such as ignoring statistical characteristics,high dimensionality of features and large amount of computation,a spatial steganalysis based on image content complexity is proposed.This method can describe and characterize the complexity of the image content by extracting the co-occurrence matrix of different directions as the classification feature vector,and increase the statistic difference between the carrier image and the steganographic image by effectively improving the existing detection feature.Experimental results indicate that this new method can improve the detection accuracy of steganography.In short,image content and steganalysis are closely related.With the continuous development of theoretical research,achievement in this area will have a profound impact on information security.