CNN-Keypoint Based Two-Stage Hybrid Approach for Copy-Move Forgery Detection

Authenticating digital images poses a significant challenge due to the widespread use of image forgery techniques, including copy-move forgery. Copy-move forgery involves copying and pasting portions of an image within the same image while applying geometric transformations to make the forged image appear genuine. Furthermore, additional processing techniques such as additive noise scaling, JPEG compression, and rotation can be employed to further conceal evidence of forgery. These factors contribute to the complexity of detecting and verifying the authenticity of digital images. The proposed work uses a combination of the CenSurE keypoint detection and a CNN architecture to detect and localize copy-move forgery in digital images. The use of CNN architecture allows the algorithm to update its learning via training data repeatedly, making it a data-driven approach. By combining keypoints with CNN features, the proposed approach can enhance the detection of copy-move forgery even in the presence of attacks such as geometrical transformations, scale, and rotation. Additionally, the proposed approach can effectively handle post-processing operations such as JPEG compression, additive noise, image blur, colour reduction, brightness change, and contrast adjustment. One important aspect of the proposed approach is its ability to handle images with different textures, including smooth and self-similar structural images with dense textures. The proposed approach can produce stable results in images with various attacks, making it a functional and reliable tool for detecting copy-move forgery in a diverse range of forged images. The proposed approach represents an important contribution to the field of multimedia forensics, providing an effective and reliable means of detecting and localizing copy-move forgery in digital images.