A Robust Hybrid Multi-Scale Approach to Detect Copy-Move Forgery in Digital Image

With the development of cyber security and multimedia forensics, digital image manipulation has recently been recognized as one of the major challenges in forensic image analysis. Therefore, selecting an image area and then copying and pasting it into the same image is the hardest process in passive image forgery. This act violates privacy and secrecy of authenticity of digital image. The attacker exploits the available tools of editing image program to make the fake image similar to the original one. This paper presents a proposed fast and efficient passive Copy-move forgery detection scheme. Hessian- Affine and Harris-Affine detectors, and Shift Invariant Feature Transform (SIFT) descriptor, are employed in the proposed scheme. These detectors provide sufficient key points for detecting the duplicated regions in the case of small or invisible regions. The experimental results show that the proposed scheme is invariant against simple and hard attacks like uniform or non-uniform transformation. The proposed scheme was evaluated using standard data sets (GRIP, MICC 220, and F8 Multi). Resulted True Positive Rate (TPR) was 0.98 and False Positive Rate (FPR) was 0.035. Thus, the scheme is effective and providing valuable results compared to recent passive image authentication schemes.