A Scoping Review of GAN-Generated Images Detection

The usage of Generative Adversarial Network (GAN) architectures has given anyone an ability to generate an image that is indistinguishable from the real image. The improper use of GAN-generated images may lead to serious privacy, security, political, and social consequences such as spreading of fake information and legal issue. Therefore, it is crucial to emphasize the widespread of fake imaginary by developing a fake image detection system. Convolutional Neural Network (CNN) is traditional method in detecting GAN-generated images. However, due to the advancement and variations of GAN, CNN often suffer from limited generalization. Benford’s law can also be applied to produce features that can be used to detect GAN-generated images. In this paper, the fundamentals of GAN, and the technology used in fake image detection model will be discussed and reviewed thoroughly.