Study On Steganalysis For GIF Images

With the rapid development and wide application of Internet communication technology, how to ensure the security and the reliability of Internet communication has become an important issue of today’s information age. Steganography, which embeds secret messages into multimedia data in a way that is statistically undetectable, is common secure communication technique. However, steganography is a double-edged sword, and illegal implementations of steganography will grievously endanger public security and social stability. Steganalysis attempts to determine whether the given multimedia data contains secret message or not so that defeat the goal of steganography. As the counter-process of steganography, steganalysis is playing an increasingly important role in information security field.The graphics inter-change format (GIF) image is one of the most suitable cover due to its popularity in Internet. Many steganography algorithms and steganography tools for GIF images are proposed. However, little attention has been paid to the steganalysis for GIF images. In this paper, GIF image steganalysis technique has been studied because of the importance and necessity of GIF image steganalysis. The main work of this paper includes the following respects:1) A steganalysis algorithm based on colors-gradient co-occurrence matrix (CGCM) is proposed. To reflect the influence of embedding secret message on color-correlation between adjacent pixels and image texture effectively, the concept of CGCM is proposed. CGCM is constructed with colors matrix and gradient matrix of the GIF image and27-dimensional statistical features of CGCM are extracted. Support vector machine (SVM) technique is adopted combing with the27-dimensional statistical features to accomplish GIF image steganalysis. Detection performance and time efficiency experiment results indicate that the proposed algorithm is more effective than other related algorithms for several existing GIF steganography algorithms and steganography tools which are popular in the Internet. At the same time, the time efficiency of the proposed algorithm is higher than other related algorithm.2) A steganalysis algorithm for GIF images is proposed based on differential zero coefficients (DZC) and index co-occurrence matrix (ICM). Firstly, the DZC and ICM are calculated and36-dimensional statistical features are extracted. Because the DZC and ICM features are sensitive to the color correlation between adjacent pixels and the breaking of image texture, the extracted corresponding36-dimensional statistical features are used to detect hidden message in GIF images. Theoretical and experimental results indicate that, compared to orther related algorithms, the proposed algorithm has a better performance. Furthermore, the sensitivity of the extracted features is analyzed and the number of training images’ influence on detection performance is discussed in this paper.3) The steganalysis algorithm using generalized difference image and color correlogram, which is proposed by Zhao et al, is improved in this paper. Aiming at the weakness of Zhao et al’s algorithm, the color-pair matrix (CPM) is computed based on the structural characteristic of GIF image.36-dimensional statistical features derived from CPM are extracted and9-dimeonal generalized difference image (GDI) features of Zhao et al’s algorithm are kept. Experimental results show that the algorithm brings great improvement on detection accuracies for multibit assignment steganography (MBA) and EzStego and the time efficiency compared with Zhao et al’s algorithm.