Fake Image Detection Utilizing Transfer Learning-Based Vision Transformer

In recent years, the creation of manipulated images has increased due to the widespread use of machine-learning techniques. Fake images generated by Generative Adversarial Networks (GANs) have led to security breaches. Many conventional image classification methods, such as Convolutional Neural Networks (CNNs), have been employed for detecting synthetic images but often struggle to address complex and varied manipulations. In this study, we propose an innovative method for identifying fake images by utilizing transfer learning in conjunction with Vision Transformers (ViT). This new architecture has shown exceptional performance in computer vision tasks. In this paper, we evaluate our approach using an image dataset comprising authentic and manipulated images. The model successfully catches complex patterns by fine-tuning a pre-trained ViT on a dataset of real and counterfeit images. The proposed transfer learning-based ViT model utilizes self-attention mechanisms for image identification, significantly outperforming CNN-based models in detecting fake images. We demonstrate that our proposed model achieves an accuracy of 99% on the test dataset. In conclusion, this study leverages the capabilities of transfer learning-based vision transformers for detecting fake images.