A ConvNet Based Procedure for Image Copy-Move Forgery Detection

Abstract

Copy-Move forgery is one of the popular image tempering procedure. In which the forger modifies the original image by creating multiple instances of some objects within the image itself. Recently, several deep convnet methods have been applied in the classification of images, forensic images, image hashing retrieval, and so on, showing better performance than the traditional method. In this paper, a new architecture of deep learning is proposed for image copy-move forgery detection. This architecture includes VGG16 as the first layer, then RPN for proposing a set of regions and an “object” score for each region and RoI for finding interested area. Both original and forged area in this method is being localized. Through intensive experiments on multiple datasets, we demonstrate that the proposed model is very effective and robust against a number of known attacks. Which outperforms other state-of-the-art image copy-move forgery detection approaches.