Publication

Abstract : In recent years, deep learning has made significant strides, especially in computer vision applications and, more specifically, in information forensics. On the other hand, data-driven approaches have shown much promise in identifying manipulations in images and videos. However, most forensic tools ignore deep learning in favour of more traditional methodologies. This article thoroughly analyses the current state-of-the-art methods for detecting and localizing image alteration using classical and deep learning-based algorithms. In addition, this review includes the latest developments in the digital image forensics field, including Convolutional Neural Networks (CNNs), while incorporating insights from classical approaches and machine learning models. Furthermore, the most significant data-driven techniques to address the issue of image manipulation detection and localization are presented and segregated into four subtopics: copy-move, splicing, object removal, and contrast enhancement. This study provides an exhaustive and up-to-date survey of the field for researchers and practitioners working in this domain. In addition, it covers the current challenges and future directions in deep learning for image manipulation detection and localization. Finally, this review’s discussion of relevant approaches and experiments will aid future exploration and development in this field.