This research introduces a novel zero-shot approach for detecting face-morphing attacks that compromise security systems, without requiring labeled training data.

• Combines multi-modal LLMs with general vision models for enhanced detection capabilities
• Achieves state-of-the-art performance on multiple morphing attack benchmarks
• Provides human-understandable explanations for detection decisions
• Demonstrates strong generalizability across unseen morphing techniques

This advancement is critical for securing border control systems and identity verification processes by detecting sophisticated identity fraud attempts without extensive training data requirements.

ChatGPT Encounters Morphing Attack Detection: Zero-Shot MAD with Multi-Modal Large Language Models and General Vision Models