Deep Fake Image Detection using Xception Architecture

Deep Fake technology has become increasingly sophisticated, posing a significant challenge to the integrity of digital content in today’s information age. This research paper introduces a novel approach in detecting deep fake images and videos utilizing the Xception model, a deep CNN known for its exceptional image classification capabilities. The study begins by collecting a diverse dataset containing real and deep fake images and videos across various domains. The Xception model is fine-tuned for deep fake detection through transfer learning. To improve model robustness, a range of data augmentation strategies and regularization techniques are incorporated during training. The results of the proposed model demonstrate an accuracy of 93.01% on the test dataset. The proposed approach exhibits promising potential in combating the spread of deceptive and malicious deep fake content across the internet and social media platforms. In conclusion, this research paper presents a reliable and efficient method for deep fake image and video detection, leveraging the Xception model’s capabilities.