Copy-move Image Forgery Detection and Localization Using Alteration Trace Net

Abstract

The prevalence of social media as a modern substitute for conventional news sources has led to the rise of fake news, which usually uses tampered photographs. This trend is frequently brought on by the rapidly falling cost of high-tech cameras and cell phones, which encourage the fast creation of computerized images. The ease of manipulating digital images has made image forgery a common worry. The volume of altered photos shared daily has greatly increased due to the quick development of commercial image altering programs like Adobe Photoshop. This phenomenon has detrimental effects, diminishing reliability and producing false beliefs in many real-world applications. This paper suggests a deep learning strategy for detecting copy-move forgeries in digital images. Here, we employ two deep learning models to identify copy move fraud in digital photos, namely Buster Net and Alteration Trace Net. The CoMoFoD dataset is used to assess the performance of the two models. The experimental results demonstrate that the Alteration Trace Net model outperforms the Buster Net model with 98.6% accuracy in identifying forgeries in photos, compared to the Buster Net model's accuracy of 96.9%.