Detection of GANs)-generated images

Rapid advances in deep learning have increased the effectiveness of deep learning algorithms such as Generative Adversarial Networks (GANs) in fake content generation. This situation brings about the difficulty of distinguishing fake and real face photographs. This study focuses on the development of deep learning-based models for the detection of fake face photographs generated by StyleGAN models. For this purpose, a total of nine different Convolutional Neural Network (CNN) models focusing on Eye, Nose and Mouth regions were developed using fake face photographs generated by StyleGAN1, StyleGAN2 and StyleGAN3. These models showed significant success in detecting fake photographs and revealed the performance variability in the detection of facial features generated by different versions of StyleGAN models. The findings of this study show how advances in deep learning can play a critical role in solving challenging problems such as fake content detection and prevention. In addition, this study emphasizes the importance of conducting in-depth analysis on how different StyleGAN models process facial features and developing new approaches in this field.