Intelligent Deepfake Detector Using Audio-Visual Clues

Deepfake media is growing rapidly and causing significant harm. Bad actors now use AI to create fake videos that appear increasingly realistic. Traditional detection tools often fail because they analyze audio or visual signals in isolation. This paper introduces an intelligent Deepfake Detection system that addresses this limitation through a novel Multi-Modal Dispersion Framework. The system identifies subtle inconsistencies by tracking how lip movements align with speech patterns. By projecting these features into a shared latent space, the model quantifies the semantic divergence between modalities. A transformer module then captures cross-modal context to detect fine-grained manipulation artifacts. Evaluated on the DFDC and FakeAVCeleb datasets, the system achieves 94.3% accuracy, demonstrating strong potential for real-time deployment. This framework provides a reliable approach to media authentication and contributes to advancing AI safety.