This project develops a deepfake detection system for both voice and facial recognition, identifying manipulated biometric data using advanced feature extraction and machine learning. Key methods include facial landmark detection, optical flow, and texture analysis to detect inconsistencies in video (using datasets like DeeperForensics-1.0), while PNCC and VGGish embeddings analyse audio. Privacy is safeguarded through differential privacy, federated learning, and homomorphic encryption to protect sensitive data. This system is robust, secure, and suitable for applications requiring high data privacy standards.