Data hiding aims to hide secret information in digital media. In contrast, steganalysis is to analyze multimedia data or other cover data that can be used for data hiding, and detect the presence of hidden messages, so as to block suspicious covert communications. In recent years, with the development of digital multimedia and network technology, information security has become an urgent problem, and steganalysis has received much attention both in theoretical and industrial fields. However, as a relatively new research field, steganalysis has many problems to be solved. In this paper, we made extensive research in blind steganalysis in images, and the contributions in this paper are as follows: 1.Proposed a novel steganalysis method based on statistical analysis of empirical matrix. This method extracts high-order moments as features based on statistical analysis of empirical matrix of image, and utilizes Support Vector Machine to train classifiers to discriminate normal images and stego images. Experiments show that this method can attack most typical data hiding schemes, and it performs better than existing methods in the art. 2.Based on the above method, we further proposed a multi-model based steganalysis framework which flexibly combines various classifiers aiming at different data hiding schemes together and optimizes each classifer. This framework can make blind steanalysis for large variety of images with high precision. Additionally, we made some tentative research for the diversity problem which has innovative sense in this field. 3.We developed two demo systems to implement the methods we’ve proposed: Data Hiding and Recovery Toolbox, and Steganalysis Demo Platform.
修改评论