The Study Of Extracting Attack On Image Steganography And Reversible Data Hiding In Encrypted Images

Steganalysis has become an important means of maintaining social stability and safeguarding national security. C urrently, there have been numerous excellent results on steganalysis, but the research mainly focuses on detecting the existence of hidden information. To extract the hiding information(extracting attack) is a research orientation of steganalysis need urgent developing, and plays a decisive role on obtaining the valid evidencesin the electronic debate in computer forensics. The present paper did some research on the extracting attack to image steganography from two aspects of payload location and stego-key recovery. Based on the result of extracting attack, this paper also studied the RDH(reversible data hiding) method in encrypted images. This is because RDH is widely used in medical imagery, military imagery and law forensics. And the RDH in encrypted images can protect the privacy of the image content besides it can act as a RDH method. So the RDH in encrypted images have a wider prospect.The specific research results are as follows:1. A method of locating the embedding payload for LSB replacing. First, we analysis the impact on WS residual made by the embedding process, and get the probability distribution of WS residual in a single image. Then we estimate the modified pixels by observing the value of the real WS residuals and the probability distribution. Last we put forward a method to locate the embedding payload when we have an access to a number of images with same size and same payload. Experimental results show that the proposed method works well, and outperforms the exsiting method in three aspects: 1)moreprecise for low embedding rate; 2) stable result for any image;3) easy to conduct with a lower computational complexity.2. A method of stego-key recovery for LSB steganography. First we analysis different generation models of embedding path and study the impact of different steganographic path generating model to the result of stego-key recovery. Then we compute the necessary number of samples and the appropriate threshold on the basis of modified pixels have been estimated, so as to reduce the computational complexity of the extracting attack. Finally, a method of extracting attack for LSB steganography is proposed and we point out that which sort of path generating mode is suitable for the method. In this paper, we give experimental results for three cases : the cover image is known, the cover image is used for double times and the JPEG decompressed image is used as cover image. And the experimental results show that: the proposed method has a wider scope of application.3. A novel method of RDH in encrypted images. This method reserves embedding room before encryption and realize the complete separation of data extraction and image recovery in operation. First, we research the feasibility of RDH in JPEG decompressed image blocks. Then we calculate the theoretical probability of recovery of decompressed image blocks successfully. Finally, a method of RDH in encrypted images is proposed and the main steps of the method include pretreatment, encryption, data embedding, data extraction and image recovery. Experimental results show that: with respect to the three existing RDH method in encrypted images, this method has less error in image recovery, and for given embedding rates, the PSNRs of decrypted image containing the embedded data are significantly improved.