2020/09/10 by Abdul Muntakim Rafi, Rafi, Abdul Muntakim, Q. M. Jonathan Wu +3
Computer Science · #AI in cancer detection #Digital Media Forensic Detection #FOS: Electrical engineering #Generative Adversarial Networks and Image Synthesis #Image and Video Processing (eess.IV) #Signal Processing (eess.SP) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2009.05379
openalex publication_date 2020/09/10 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Source camera model identification (CMI) and image manipulation detection are\nof paramount importance in image forensics. In this paper, we propose an\nL2-constrained Remnant Convolutional Neural Network (L2-constrained RemNet) for\nperforming these two crucial tasks. The proposed network architecture consists\nof a dynamic preprocessor block and a classification block. An L2 loss is\napplied to the output of the preprocessor block, and categorical crossentropy\nloss is calculated based on the output of the classification block. The whole\nnetwork is trained in an end-to-end manner by minimizing the total loss, which\nis a combination of the L2 loss and the categorical crossentropy loss. Aided by\nthe L2 loss, the data-adaptive preprocessor learns to suppress the unnecessary\nimage contents and assists the classification block in extracting robust image\nforensics features. We train and test the network on the Dresden database and\nachieve an overall accuracy of 98.15%, where all the test images are from\ndevices and scenes not used during training to replicate practical\napplications. The network also outperforms other state-of-the-art CNNs even\nwhen the images are manipulated. Furthermore, we attain an overall accuracy of\n99.68% in image manipulation detection, which implies that it can be used as a\ngeneral-purpose network for image forensic tasks.\n