Merging Feature For JPEG Image Universal Steganalysis

The Internet has gradually become a very important communication channel since1900s.People often send and receive information via the Internet, so we must protect the information on the Internet. Information hiding is an invisible communication technology, which can hide important information in the other vector. So information hiding technology can conceal the existence of hidden information. On the other hand, with the development of information hiding technology, we face some serious security risks. As a consequence, we must find and detect hidden information on the Internet to protect information and avoid illegal abuse.In recent years, the JPEG image format is increasingly popular, and it has been widely applied. Therefore, research on JPEG image format steganalysis is becoming more and more popular.In this paper, we mainly discuss about information hiding and steganalysis technology for JPEG image format, including method theory, method process and their limitation. On this situation, this paper proposed a new steganalysis method for JPEG images based on merging feature. This method merged some excellent feature. In this paper, we call this method JFMP-355. The experimental results show that this method can effectively detect the F5,JSteg,OutGuess,MB1that is based on8×8block, and this new method improve the detection accuracy of the existing steganalysis methods. However, this method can not detect YASS.YASS (Yet Another Steganographic Scheme) is a new information hiding technology, which can be resistant to several high detection rate of steganalysis methods. After determining the big block, YASS selects sub-block randomly. As a consequence, YASS is an information hiding method that is not based on8×8block. Because of its random selection of sub-block, YASS has a high degree of security. In this paper, we proposed a new universal steganalysis method for JPEG image, which can detect YASS. On the basis of JFMP-355method, the new method modify the calculation of feature. Difference calculation is replaced by Cartesian calculation. In the end, we can get710dimension feature for steganalysis and we call this method JFMP-710. In the experiment part, we set different bock size, different embedding rate and different quality factor of JPEG images as test image library. Compared with the other steganalysis method, the experimental results show that the proposed method has a higher detection rate of F5,JSteg, OutGuess,MB1and YASS, and it is an universal steganalysis method.