Providing a Method to Process Images for Detecting Image Fraud Using a Decision Tree Technique

Digital image forgery detection techniques are collection of algorithms to detect forgery, regardless of tools and complexity degree. Their common aim is to detect forgery in digital images. Today, there are many powerful software tools for manipulating digital images. The exchange of documents and information electronically, regardless of the security infrastructure, is a threat to Information and Communication Technology (ICT). The decision tree technique is a common method for categorizing images, because they are made easily by several nodes that can easily be interpreted by designers. In this study, we have tried to differentiate between the original and the forged image through a careful examination. We performed this technique on 50 manipulated images using a main algorithm (decision tree) and two auxiliary algorithms of Scale Invariant Features Transform (SIFT) and Random sample consensus (RANSAC) that had been merged. The comparisons between algorithms showed that efficiency, reliability and time complexity degree have been effective in the main algorithm using the conversion of image pixels to the numerical matrix, the formation of the corresponding graph for each image and the comparison of the formed graphs. In the present study, the main algorithm could detect 86% in 162.59 seconds while the auxiliary algorithm could detect 82% in 205.92 seconds.