On the reliability of forensic schemes using resampling for image copy-move forgery

摘要

The goal of image forgery detection in multimedia security and forensics is to find a clue of image manipulation and prove the inauthenticity of digital images. The presence of the original and its duplicated regions in an image is regarded as a fingerprint for copy-move forgery. The problem investigated here concerns the situation when resampling process is employed in the pre-processing stage of detecting image copy-move forgery. In several detection techniques of copy-move forgery in digital images, resampling is utilised to lower the spatial resolution of an image, further improving the efficiency and the speed of image forensics. However, the reliability of detection methods has not been examined in detail. In this paper, we take a view of some recently-proposed forensic techniques using resampling and analyse the reliability of detecting copy-move forgery, by modelling copy-move forgery process and detection problem. Based on theoretical analysis and experimental validation it is concluded from this study that a correct detection of image copy-move tampering may be impeded by resampling operation under certain conditions and it is more difficult to reveal the forgery than previously thought.