Detection and localization of copy-move tampering along with adversarial attack in a digital image

Detection and localization of copy-move tampering along with adversarial attack in a digital image

연구 분야: Analysis

학회: Discover Computing

Authenticating digital images is increasingly challenging due to the prevalence of tampering techniques such as copy-move tampering, where parts of an image are copied and pasted within the same image. This tampering is often disguised using geometric transformations like rotation and scaling, and further concealed by techniques such as JPEG compression and AWGN. In this paper, we propose a novel approach for copy-move tampering detection that leverages the SURF detector and BRISK descriptor. The SURF detector is known for its speed and stability, even in rotated images, and when combined with BRISK, it significantly improves the F1-Score compared to existing approaches. Our approach employs hierarchical clustering and neighborhood search to accurately identify and locate tampered regions, even in images that have undergone post-processing techniques such as geometric transformations, combined attacks, and multiple instances of copy-move tampering. We demonstrate that our approach outperforms existing keypoint-based approaches, particularly in scenarios where images have been subjected to complex manipulations like rotation, scaling, AWGN, and JPEG compression. Furthermore, considering execution time, our approach holds promise for real-time copy-move tampering detection and image authentication applications. This improvement in both detection accuracy and speed makes our approach a valuable contribution to the field of digital image forensics.