1. WASIN ALKISHRI - Faculty of Communication, Visual Art, and Computing, University Selangor (UNISEL), Selangor, Malaysia.
2. Dr. SETYAWAN WIDYARTO - Faculty of Communication, Visual Art, and Computing, University Selangor (UNISEL), Selangor, Malaysia.
3. Dr. JABAR H. YOUSIF - Faculty of computing and information technology, Sohar University, P.O. Box 44, Sohar, Oman.
4. MAHMOOD AL-BAHRI - Faculty of computing and information technology, Sohar University, P.O. Box 44, Sohar, Oman.
Deepfakes are a type of "artificial intelligence" that involves the use of genuine images or videos that are then transformed into false forms of media for a particular goal. Deep learning algorithms, implemented in software, are used to accomplish this goal. As deep generative models like generative adversarial networks can be visually indistinguishable from real photos, their potential harmful application raises concerns, such as annoyance, embarrassment, provocation, terrorism, extortion, falsification of information, and intimidation. Because of this, industry and governments have become increasingly concerned about distinguishing between them and limiting their use. In this paper, we present an analysis of the highfrequency Fourier transform model of real and deep network-generated images and show that deep network-generated images include some unreal properties, even if these properties are not obvious to the human eye. In order to determine the most effective model to distinguish between original and fabricated images, frequency domain analysis will be applied to two classifiers, custom VGG16 and Dense Net-121. The goal of this study is to evaluate the effectiveness of our technique with the use of the 140 k Real and Fake Faces datasets of deep fake image. The findings of our experiments indicate that the difference between spectra in the frequency domain is a practical artifact that can be used to efficiently recognize different kinds of GAN-based generated images.
Discrete Fourier transform, VGG16 and Dense Net-121, Deepfakes.