The Blind Forensics Research On JPEG Image Copy-move Tampering

With the advent and growing popularity of digital multimedia recorders, PCs and image editing software, tampering digital image is becoming increasingly easy. If the falsify information was being abused, great impact on the personal privacy, social development and political stability would occur. Therefore, the research on digital image forensics is very important.JPEG is the main kind of digital image formats, there are many tampering methods about it, the tampering based on copy-move of JPEG image is the most common. This paper focus on JPEG image forensics, the main work is as follows:(1) A method based on block artifacts for detecting copy-move tampering on the same JPEG image is proposed. Through analyzing the process of JPEG compression, the block artifact caused by copy-move operation in the JPEG image is found. By utilizing the characteristics of the block artifacts, a method based on block artifacts for detecting copy-move tampering is proposed. This method can detect the copy-move tampering without artificial intervention. Experimental results show that the proposed method not only has good detection accuracy, but also has strong robustness. Even tampered image is suffered from JPEG compression, adding noise or cutting processing, the algorithm still has a good detection result.(2) A method based on quantization error for detecting copy-move tampering between the different JPEG images is proposed. Copy-move tampering between the JPEG images with different quantization factors will not only introduce the block artifacts, but also generate quantization error. By utilizing the characteristics of the quantization error, a method based on quantization error for detecting copy-move tampering is proposed. Compared with existing algorithm, the proposed method not only greatly raised the detection accuracy, but also improved the time complexity.This paper proposed detection methods based on block artifacts and quantization error, experimental results show that the methods not just have robustness, but can effectively detect the copy-move tampering between either single or multiple images.