Image Forgery Detection Using Multi-Layer Convolutional Neural Network

Abstract

AbstractImage forgery detection is among the most appealing research domains in the field of multimedia forensics. Distinguishing tampered images from pristine images has become increasingly difficult due to growing number of sophisticated manipulation techniques. Digital images can be easily tampered using copy-move or splicing techniques and the traces of forgery can be hidden using post-processing operations. The field of digital image forensics majorly aims at detecting image forgery and ensuring the authenticity of digital visual media. In this manuscript, we propose a passive approach for digital image forgery detection. It is a non-block-based functional framework that utilizes deep convolutional neural network to distinguish tampered and pristine images. To build an efficient and robust model, extensive experiments have been evaluated on COVERAGE dataset and the results suggest that it attains state-of-the-art performance with greater robustness to post-processing tampering operations. The comparative results have also been included for better evaluation.KeywordsImage forgery detectionDeep LearningCopy-move tampering detection