This research traces how facial deepfake detection has evolved from analyzing single data types to sophisticated multi-modal approaches that integrate audio-visual and text-visual signals.

Key Developments:

• Presents a structured taxonomy of detection techniques across different modalities
• Analyzes the transition from GAN-based to diffusion model-driven deepfake generation and detection
• Examines integration of audio-visual and text-visual cues for more robust detection
• Addresses challenges in combating increasingly realistic synthetic media

Security Implications: As synthetic media becomes virtually indistinguishable from authentic content, this research provides critical frameworks to combat identity fraud, misinformation campaigns, and social manipulation through advanced detection methods.

Evolving from Single-modal to Multi-modal Facial Deepfake Detection: Progress and Challenges