Research On Passive Forensics For Spliced Image

ABSTRACT:Digital image is becoming a more important role in information dissemination. At the same time, Digital image can be forged easily with the development of digital image processing technology. The authority of digital image is challenged. The security of digital image content can be protected by the technology of digital image forensics,so it became an important information security technology.Recently the passive digital image forensics is developing rapidly, which based on image itself. Compared with other image authentication approaches, passive techniques for image forensics operate in the absence of watermark or signature, it have better usability. In this paper, the focus is the passive detection technology for spliced image which is the most common tamper mode. At first, Authentication model of spliced image and its system solution is be proposed on the results of previous research. There are three items in image splicing forensics, which are:(1)Classifying on the image whole feature;(2)Locating the spliced position (3) Detecting the post-processing traces in spliced image. The results of the paper are as follows:(1) Spliced image classification on its feature——Based on the characteratic of spliced image, we extracted features from two aspects:one is the correlation of pixel which was obtained by Markov transition probability matrix of the pixels bit-plane; the other is the high frequency characteristics of the difference between of color components, these characteristics can be extract by wavelet transform. Then we classify spliced image and nature image on these features, and the classification accuracy is high. In this process, the number of features is smaller and the complexity of features is simpler.(2)Locating the spliced position——Locating the spliced content is a necessary step of spliced image forensics. Two algorithms are proposed for this problem in the paper:By the first algorithm, we can get the difference between pixel value and CFA interpolation, and the position of imge splicing can be revealed by the distribution characteristic of the difference. In the second algorithm, Background noise of spliced image are be analyzed, There are different noise feature between spliced part and original part in spliced image, so the splicd part can be found. The spliced part can be located exactly by these two algorithms.(3)Detecting the post-proessing traces——In order to improve the robustness of spliced image forensics, we proposed two algorithms to detect post-processing traces in spliced image:Using the first algorithm, the blured traces can be found by the feature of edge; Using the second algorithm, the recaptured image can be detected by the light direction. These two algorithms can detect the trace of blured and recaptured in post-processing.Finally,the paper is summarized and the prospective developing direction of spliced image forensics is also put forward.