Research On Extracting Attack To Hidden Message In Image

Steganography is the art and science of covert communication and it can be applied inmilitary communication and illegal activities. As the technology of detection, extraction anddestruction of hidden message, steganalysis is signicant to prevent illegal abuse of steganography.Extracting attack is the ultimate purpose of steganalysis. It can obtain crucial information, so it isof great importance for military initiative and information security. In this dissertation, somegroping researches on the theory and methods of extracting attack are done. The maincontributions are summarized as follows:1. The attack model is built by analyzing attacking motive, and then the idea and categoryof extracting attack are analyzed and the problem of actual research on extracting attack isilluminated.2. The difficulty of extracting hidden message is studied. The Random Oracle Model is usedto formalize the model of extracting attacker, then the difficulty model is built and the attackingpoint for extracting attacker is found. By describing the course of extracting attack, theextracting difficulty is obtained, and some general principles for devising extracting attackmethods are proposed by analyzing the extracting difficulty.3. Extracting attack method on random replacing steganography is researched. Based ongeneral principles, the extracting attack can be recognized as a special kind of detection. Theimage pixels are divided and the impacts replacing embedding and least significant bit flippinghave on different pixels are analyzed, and then an obvious statistical difference used fordistinguishing correct key is obtained and an effective extracting attack method is presented.Experimental results show that this method can extract hidden message successfully in a certainrange of embedding ratio.4. Extracting attack method on sequential LSB steganography is brought up. According togeneral principles, the stability of embedded part and the growth of un-embedded part areanalyzed in theory, and an obvious statistical difference used for detecting correct key is got. Byestimating the location of stable part, the effective key space is reduced, and then, according tothe obvious statistical difference, an extracting attack algorithm with good extractingperformance is proposed.Finally, the researching work is summarized, and further research topics and directions areprospected.