Detection and Identification of Fake Images using Conditional Generative Adversarial Networks (CGANs)

Abstract

Fake photos spread through social networks to a growing speed of light Anybody can use business media editing tools to edit or generate fake images - removing, adding and cloning people & objects. A number of effective approaches have been recommended to recognize these common fakes, but new attacks appear every day. Image-to-image transferral using the conditional generative adversarial networks (CGAN) looks like one of the nastiest out there as it could change image context and explanation in a very gruesome way. The study, carried out on a dataset of 2661 real images and 2661 generated images taken from UC Berkeley's official directory. Our proposed approach obtained an accuracy of up to 99.99% in detecting fake images using CGAN. Our methos is highly effective and robust when compared to existing techniques.