GAN Generated Fake Human Face Image Detection

In recent years, Generative Adversarial Networks (GANs) have revolutionized the generation of synthetic data that closely mimics real-world distributions. This research paper focuses on detecting fake human face images generated through GANs. The paper provides a thorough analysis of the current state of GAN-generated fake human face detection and proposes a novel method for robust detection. Existing detection methods often struggle with newly emerging GAN architectures, lack generalization capabilities, and are prone to adversarial attacks. In this paper, authors propose an efficient Convolutional Neural Network (CNN) architecture that detects StyleGAN3-generated fake human faces. To enhance the robustness of the model the algorithm employs a series of filters to extract image data, performs grayscale normalization and convolutional operations to find out whether the images are fake or real with more accuracy. The outcomes of the experiments demonstrate that the approach outperforms the current systems in terms of robustly identifying fake images. Authors achieved an accuracy of 99.42%. This system can be integrated into social media platforms to identify fake profile pictures or deepfake images that are often used for impersonation or spreading misinformation.