Next-Gen Detection: The Role of Vision Transformers and Statistical Methods in Autonomous Deep Fake Authentication

Abstract: Deepfake technology undermines the authenticity and integrity of digital media, requiring the creation of advanced autonomous detection systems. This study investigates the amalgamation of Vision Transformers (ViTs) and statistical artifact analysis to establish a resilient deepfake authentication framework. ViTs' capacity to extract global characteristics is combined with other statistical techniques in the suggested strategy to achieve high detection accuracy, generalisation, and attacker resistance. Tests on two benchmark datasets, Face Forensics++ and DFDC, demonstrate superior performance compared to previous convolutional models. Future studies on autonomous deep fake identification will focus on transparency, equity, and scalability.

Keywords: DFDC++, Face Forensics++, Visual Transformer, statistical analysis, autonomous authentication, and deep fake detection.