A Content Independent Digital Image Forensics Framework for PRNU-Based Image Forgery Detection and Localization

Abstract

The rapid proliferation of advanced image editing software has made digital image forgery increasingly prevalent, posing significant threats to the integrity of digital images used as evidence in forensic investigations, legal proceedings, journalism, and social media. This paper proposes a passive copy-move forgery detection technique, which relies on camera fingerprints called Photo Response Non-Uniformity (PRNU), without being dependent on the image content. A camera fingerprint is caused by manufacturing variations and other imperfections in the camera sensor, and is specific to each camera and will be present in every image taken by the camera. The method proposed in this paper is based on the fact that in the input image, the spatial domain averaged frames are extracted from the camera fingerprint and the Normalized Cross-Correlation (NCC) is calculated to determine if the image is tampered or not. Two practical situations are considered: (1) given both the original and forged images, and (2) given only the candidate image (which was captured from the same source camera) and the forged image. To localise the forgery, fingerprint subtraction, morphological dilation and thresholding is done to accurately identify the tampered area. Experiments are done on a set of 500 images where the proposed technique has been proved to be 100% successful in Scenario 1 and 97% successful in Scenario 2 when tested with several state of art CBIR techniques. The results prove the reliability and efficiency of the camera fingerprint based approach in the passive image forensics.