AI Based Deepfake Detection

This project aims to develop anAI-based system for detecting deepfake videos addressing the growing threat to digital security andmedia authenticity. The system works by detecting faces, selecting important key frames, and extracting visual features using a Large Vision Model (Contrastive Language–Image Pretraining).It further analyzes temporal inconsistencies through an LSTM network combined with an attention mechanism. The proposed approach achieves up to 95% accuracy and offers practical applications in social media monitoring, digital forensics, and the prevention of misinformation.

Poster