Image Tampering Detection Using Convolutional Neural Network

Abstract

Pictures are considered the most reliable form of media in journalism, research work, investigations, and intelligence reporting. With the rapid growth of ever-advancing technology and free applications on smartphones, sharing and transferring images is widely spread, which requires authentication and reliability. Copy-move forgery is considered a common image tampering type, where a part of the image is superimposed with another image. Such a tampering process occurs without leaving any obvious visual traces. In this study, an image tampering detection method was proposed by exploiting a convolutional neural network (CNN) for extracting the discriminative features from images and detects whether an image has been forged or not. The results established that the optimal number of epochs is 50 epochs using AlexNet-based CNN for classification-based tampering detection, with a 91% accuracy.