Copy-move forgery detection based on deep learning

Abstract

Copy-move forgery detection (CMFD) is probably one of the most active research areas within the blind image forensics field. Among existing algorithms, most of them are based on block and key-point methods, or combination of them. Recently, some deep convolutional neural networks methods have been applied in the image classification, image forensic, image hashing retrieval, and so on, which have shown better performance than the traditional method. In the work, a novel copy-move forgery detection method based on convolutional neural network is proposed. The proposed method uses existing trained model from large database as ImageNet, and then adjusts slightly the net structure using small training samples. Experimental results show that the method we proposed obtains satisfactory performance to the forgery image generated automatically by computer with simple image copy-move operation.