Digital Image Forgery Detection using Deep Learning

In recent years, the rise in accessibility and simplicity of image editing tools has led to a surge in the production and dissemination of fake or manipulated images across media and the internet. Numerous techniques have been introduced to verify the originality of images, for certain cases, to pinpoint the specific regions that have been altered or forged. This paper offers a review of several of the latest image forgery detection methods, particularly those based on Deep Learning (DL) frameworks, with a focus on prevalent attacks such as copy-move and splicing. The growing phenomenon of Deepfake content is also discussed, especially in its application to images, which results in effects similar to splicing. The relevance of this survey is highlighted by the fact that deep learning-based methods currently offer some of the best performance on standard datasets. We explore the essential features of these methods and describe the datasets used for their training and validation. Furthermore, we assess and compare their performance where applicable. Based on this examination, we suggest potential future directions for research, particularly in the areas of deep learning architectures, evaluation methodologies, and the development of datasets to facilitate easier comparison of these techniques