Forensic approach for distinguishing between source and destination regions in copy-move forgery

The copy-move forgery involves copying a semantically part from an image and pasting it into a different location within the same image to change the context of the image and deceive people. The copy-move forensic detectors authenticate the image and localize similar regions without identifying which region is original or forged. Because the forged region is replicated from another region of the same image, its characteristics have been inherited, making it difficult to distinguish between the original and forged regions. This paper proposes two approaches as a second stage, after localizing duplicated regions, to distinguish between source and destination regions. The adjacent pixels in images are non-independent and have some correlations that would be destroyed due to modifying the image. The deviation of these correlations would expose traces left due to the image forgery and is evaluated by the Joint Probability Matrix (JPM) in the first approach and by the Local Binary Pattern (LBP) in the second approach. Both approaches employ Jensen Shannon Divergence (JSD) to measure the correlation between feature vectors and generate similarity scores to distinguish between the source and the destination regions. The proposed approaches were demonstrated by employing the GRIP dataset with six post-processing operations to conceal forgery. The experiments exhibit a high accuracy rate of 96.25%.