Multi-Feature Steganalysis For Palette Images

With the age of information, individuals or institutions can transmit information easily and quickly through variety of communication networks. Because much of the secrete information is very easily stolen or damaged in the public communication platform, so information security has attractted more and more attention. As an important field of information security, steganography and steganalysis are two problems that instense interest of the governments and researches.Steganography is needed to cover the 'ongoing secret communications'. After embedding secret information in digital carriers, such as digital images, audio and etc, the secret information can be received by the receiver without being perceived. Steganalysis is an attack of steganography. In order to reveal the existence of secret information, the suspect digital carriers should be analyzed. The "ongoing secret communications".is been aware, and then the secret information will be analyzed. Palette images (such as GIF format image) which are widly used as carriers for transmiting secret information are common format images in the Internet. Many steganography softwares, such as Gifshuffle, EzStego and etc, which based on palette images, are appearing, so the research of steganalysis for palette images has practical significance.In this paper, steganography and steganalysis of palette images are mainly researched, and the main work of this paper is listed as follows:(1) Based on the simulation of Multibit Assignment (MBA) steganographic algorithm for palette images, different embedding rates of stego images have been got. By comparing and analysising the stego images and the original images, the changed degree for images the time efficiency of the MBA algorithm has been tested. The entire MBA steganography has been programmed by using C programming language, including the embedding part and the extracting part. The feasibility of assignment bits by using the list of sides and the necessity of embedding the length of secret information has been given out.(2) A multi-featrues steganalysis algorithm for MBA steganography is being proposed by analysising the effect of image characteristics out. In this steganalysis algorithm, firstly, the difference matrixs of three color planes and the ones of corresponding gray-scal plane have been calculated. And then the features of co-occurrence matrix and high-order histogram from the difference matrixs have been extracted. After reducing the dimensionality and normalizing the value of featrues, support vector machine is used to classify and detection. The experimental results show that the accuracy of detection in this steganalysis algorithm is better than Farid's algorithm in a certain degree.