Research On Digital Image Manipulation Forensics

With the rapid development of digital media editing techniques, digital image alter-ation and tampering become quite easy. The illegally manipulated and forged images emerge frequently. As such the credibility and security of digital media information are destroyed seriously. In the applications such as law enforcement and news recording, it is necessary to verify the originality and authenticity of digital images, and make clear the image manipulation history to get more information. Therefore, the research on dig-ital image manipulation forensics is significant and valuable for realizing the auth-entication of multimedia content. This thesis addresses the digital image manipulation forensics problem and gains the following novel research achievements:1) Two splicing detection algorithms are proposed based on the consistency analysis of edge related features in digital images. First, the consistency between the CFA (Color Filter Array) interpolation artifacts in natural edges and those in splicing boundaries is exploited to design an image slicing detection algorithm. Comparing with the most prior methods based on statistical learning, our proposed method is cost-efficient and owns lower algorithm complexity. Second, another splicing detection algorithm is proposed based on the estimation of edge-based blurriness. Such an algorithm could be used in the practical scenario that post-burring is enforced onto the splicing boundary, which would be located accordingly.2) Two forensic algorithms are proposed to detect the digital image USM (Unsharp Masking) sharpening and median filtering manipulations, respectively. The authors address the image sharpening and median filtering forensics problems earlier in the digital forensics community. The USM sharpening process is modeled in signal and described in mathematics. The mechanism of producing overshoot artifacts is analyzed. And consequently, an effective sharpenging detection scheme is designed. Meanwhile, the statistical abnormity in the image’s first-order difference domain, which is incurred by median filtering, is analyzed theoretically. A fast and effective median filtering detection algorithm is proposed. Test results show the proposed sharpening and median filtering forensic algorithms could identify the corresponding manipulations efficiently.3) A series of algorithms on image contrast enhancement forensics are proposed. To efficiently detect contrast enhancement in middle/low quality JPEG compressed images, a global contrast enhancement detection algorithm is proposed based on the shape an- alysis of the gray level histogram peak/gap bins. By means of the correspondence be tween the peak/gap bins position distribution and the involved pixel value mapping, a fast and efficient algorithm is proposed to estimate the gamma parameter blindly. As for the scenario that source regions suffer different contrast enhancement mappings, an effi-cient source-enhanced spliced image detection method is designed. Lastly, the security of existing contrast enhancement forensic algorithms is analyzed briefly. Test results demonstrate that our proposed contrast enhancement forensics methods could achieve good performance.4) A semi non-intrusive forensic algorithm is proposed to identify the digital image resampling operator. As for the strictly monotone signal, its first derivative polarity regularity after suffering the traditional and geometric-dithering types of resampling is analyzed theoretically. Through designing suitable test pattern images, a unified resam-pling operator identification scheme is proposed. Test results show that the proposed method could identify the resampling operator embedded in softwares, as well as detect the anti-forensic type of resampling manipulation in some specific scenarios.