Research On Anti-detection Steganographic Method Based On Image

Information hiding is a newly risen technique in the art of information security of digital multimedia. It can be applied to protecting or transmitting important information, such as texts, audios or videos. The anti-detection property of steganographic method is an important requirement of steganography. It is a necessary condition to assure the security of hidden information. The anti-detection property has an indiscerptible relationship with other technical requirements, such as embedding efficiency, embedding rate, invisibility, etc. From the angle of these requirements, this paper presents some steganographic methods to increase security. The detailed contents are as follows.After researching on the question of embedding efficiency and embedding rate, two matrix embedding methods: single bit matrix embedding and W^L matrix embedding are proposed. The characteristic of single bit matrix embedding is able to hide maximal data given the length of the cover data unit and modifying the least data, which means reach the maximal efficiency. The characteristic of W^L matrix embedding is high embedding efficiency, small changes in the host information and capable for embedding lagre amount of information.Some researches on practical applications have been done. The single bit matrix embedding method is applied into information hiding in binary images and MPEG-2 videos and achieves good results. The W^L matrix embedding method is applied to hiding a text or a speech into cover GSM speech.By analyzing the detection principle of sample pair analysis (SPA) algorithm, a steganographic scheme against SPA is proposed. In further, a SPA detection feedback scheme is proposed which the advantages, compared with other existing methods, are complete security to SPA detection, high embedding rate, low distortion to cover images and fast embedding process.In addition, combining the project of laboratory, some work on steganalysis has been done and a steganalysis method based on PCA and SVM is proposed. The proposed method can detect many popular data hiding methods with high correct detection rate.