Copy-Move Forgery Detection in Digital Forensic Images Using CNN

Abstract

Digital images are commonly used for sharing visual information and can be manipulated easily. The detection forgery in digital images has become a hot domain of research in digital image forensics due to the prevalent use of image editing tools for manipulating an image to conceal or distort information in the image. One of the most common image forgeries performed is the copy-move forgery. This type of forgery involves copying a segment of the image, which is then pasted to a different segment of the same image. The need for detecting whether an image is authentic becomes essential. The existing methods implemented for detecting image forgeries were based on traditional feature extraction algorithms such as block-based and key point-based algorithms. These traditional techniques employed produce a low-performance result. Deep learning techniques have proven to provide better performance in image processing tasks. In this research, a convolutional neural network based on a pre-trained ResNet50 network was proposed to detect copy-move forgeries in digital images. The proposed model uses the CoMoFoD image dataset in experimenting. The metric evaluation results achieved in the proposed model show that deep learning methods performance is more effective in digital image copy-move forgery detection.