Uncovering DeepFake Images for Identifying Source-images

Owing to the swift advancement of face swapping technology utilized in creating "Deep-fakes" and the increasing public awareness regarding multimedia security, numerous studies have been undertaken in recent years to devise various methods for detecting deepfakes. This paper presents a framework aimed at enhancing clarity in deepfake detection techniques through the reconstruction of source images from a provided deepfake image. Our focus is on developing a system that not only addresses the inquiry, "Is this image a DeepFake or not?" but also provides insights into the question, "If the image is a DeepFake, what are the source images?" An extensive experiment was conducted on our hypothesis, yielding significant results from the exploration.