Study Of Steganalysis Using Wavelet Higher-Order Statistics For JPEG Images

With the rapid development of Internet, the problem of network security has become more and more important. What discussed in this paper is within the scope of project,'Information Security Supervision System on Internet', which is proposed to solve the problem of content security on Internet. The task of this paper is to steganalyse images transferred on Internet, which is to say to judge if there is some secret message hidden in images. The paper focuses on those steganographic images hidden on frequency domain.In the first part, the paper introduces the development of information hidden and the background of the project. Basic theories of data hidden, including definition, classification, characteristic and scheme etc, are discussed.Then some steganography methods hiding message on spatial domain or frequency domain of images are analyzed in detail, especially to those based on JPEG images, such as JSteg, OutGuess and F5 etc.In chapter 4, the paper discusses the algorithms and characteristics of steganalysis methods in detail, which include three kinds of steganalysis methods: steganalysis on spatial domain, steganalysis on frequency domain and universal methods. Merits and shortcomings of these methods are also presented.Then this paper presents a new steganalysis method for detection of steganography in image's frequency domain, Steganalysis based on higher-order statistics of YCbCr panels for JPEG images. The detecting method is universal and not restrained to some special steganography methods. While the JPEG images are presented in YCbCr color channels, not RGB that in computers, so if the higher-order statistics are still extracted on RGB color channels, the characteristics of steganographic images are inevitably be weakened. We will analysis the reason in chapter 5. In this paper, the statistics are extracted in YCbCr color channels, and get a satisfying result.