Image Forgery Localization and Detection using Multiple Deep Learning Algorithm with ELA
Anshul Kumar Singh, Chandani Sharma, Brajesh Kumar Singh
Abstract
In today’s technical world, social media may help an individual grow significantly. On the other side, we must not overlook the reality that it is also a large platform for criticism. With recent advancements, approaches for creating and manipulating multimedia information may now deliver a highly sophisticated level of realism. There has been a blurring of the line between real media and fake media in recent years. Creative arts, film production, advertising, and video gaming are among the industries that could benefit from this technology. However, it poses significant security risks. Software tools freely accessible on the internet enable anybody, with no particular abilities, to generate extremely convincing phony images and films. A dataset composed of real photographs and false images is used in this article to identify images of modifications using Deep Learning algorithms with Error Level Analysis on each image. Our experiment yielded the accuracies of 93.5%, 89.1 and 92.4% in ResNet50, Vgg16 and CNN respectively for 50 epochs.