Deepfake Image Detection for Low and High Quality Images for Biometric Face Recognition

Deepfake technology has become a significant threat due to its potential for generating realistic fake images and videos. It has emerged as a significant concern due to its potential for misinformation and the manipulation of visual content. Detecting and mitigating deepfake content is of utmost importance to combat the spread of misinformation and safeguard the integrity of media. In this paper, we propose a scheme for deepfake detection, focusing on both low and high quality images. Our approach leverages the power of deep learning networks, specifically the XceptionNet and a GAN network comprising EfficientNet, to effectively discriminate between real and manipulated images. We conduct extensive experiments on diverse datasets, including GAN-generated and real images of different qualities, and evaluate the accuracy of the suggested models’ performance, robustness and computational efficiency.