Roadmap for Digital Image Forgery Detection Using Deep Learning

Abstract

Technology advances are prominent today while influencing all aspects of our lives. Misuse of information has also increased as a result of technical improvements. As a result, investigators have the enormous task of recognizing modified information and distinguishing this from genuine data. Among the most prevalent techniques for electronic image alteration is splicing, which includes replicating a specific section using the same or different photograph and transferring it to a new image. In the wake of this issue, picture identification of forgeries has arisen as a viable approach for confirming the validity of online photographs.

Within this paper, we propose a strategy depending on the cutting-edge ResNet50v2 neural networks framework. This study describes an approach intended specifically for detecting splicing, among the most common forms of online picture forgeries. The VGG-16 convolutional neural network model is used in our technique. The recommended network topology accepts picture patchwork as data and generates identification outcomes on the patch, categorizing them as legitimate or fraudulent. We select pieces from the primary picture areas and the inserted splicing boundaries throughout the teaching stage. This paper proposes an effective technique for identifying splices in electronic photos, proving the usefulness of our deep learning-powered strategy and emphasizing its superior results over previous alternatives.