Deepfake Detection Based on Visual Lip-sync Match and Blink Rate

Deepfake technology has emerged as a significant challenge to the authenticity of digital media, necessitating innovative detection methods. This paper introduces TrueSync, an advanced application for detecting deepfake videos by integrating two critical detection features: lip-sync analysis and blink rate monitoring. Leveraging a hybrid approach combining CNN-LSTM and SyncNet models, TrueSync processes visual and temporal features to identify anomalies in lip movement synchronization and eye blinking patterns. The application utilizes a modular pipeline to analyse these features independently and then fuses the results for a comprehensive detection score. This approach enhances detection accuracy and provides users with reliable tools to combat sophisticated manipulations. By proposing this scalable solution, TrueSync addresses the increasing difficulty in distinguishing authentic videos from manipulated content, fostering trust in digital media.