A Detection Method for Copy-Move Forgery Attacks in Digital Images

Digital images are carrying important information in many real-world applications such as surveillance, courts of law as evidence for crimes, journalism, scientific publications, and medical imaging. Manipulating these images becomes easy due to the paramount image processing and editing tools. Thus, the authentication of digital images becomes one of the principal problems in digital image forensics and attracts the attention of the research community. In this paper, we focus on the detection of copy-move forgery attacks in digital images. It is one of the most actively used attacks. It copies some region of the image and pastes it into one or more parts of the same image to cover or emphasize important scenes in the image. We propose a method to detect such forgery regions based on Speeded Up Robust Features (SURF) and Binary Robust Invariant Scalable Keypoints (BRISK) descriptors. Both fused features are matched and performed clustering to reduce the false positives using Hierarchical Agglomerative Clustering (HAC). The proposed method is tested on real-world copy-move datasets. Experimental results show that our method is robust against various geometric transformations and able to precisely determine the forged regions.