Authors: Yichen Shi, Yuhao Gao, Yingxin Lai, Hongyang Wang, Jun Feng, Lei He, Jun Wan, Changsheng Chen, Zitong Yu, Xiaochun Cao

Published on: February 06, 2024

Impact Score: 8.45

Arxiv code: Arxiv:2402.04178

Summary

• What is new: Introduces SHIELD, a new benchmark for evaluating Multimodal Large Language Models (MLLMs) on face spoofing and forgery detection, and a novel Multi-Attribute Chain of Thought (MA-COT) paradigm.
• Why this is important: Lack of exploration into MLLMs’ sensitivity to subtle visual spoof/forged clues and their performance in face attack detection.
• What the research proposes: Designing true/false and multiple-choice questions to evaluate multimodal face data in face security tasks, using three different modalities for anti-spoofing and evaluating GAN-based and diffusion-based data for forgery detection.
• Results: MLLMs demonstrate potential in the face security domain, showing advantages in interpretability, flexible reasoning, and joint detection capabilities. The MA-COT paradigm effectively aids in subtle spoof/forged clue mining.

Technical Details

Technological frameworks used: SHIELD, Multi-Attribute Chain of Thought (MA-COT)

Models used: GAN-based, diffusion-based data models

Data used: RGB, infrared, and depth modalities for anti-spoofing; visual and acoustic modalities for forgery detection

Potential Impact

Security companies specializing in facial recognition and authentication technologies, companies integrating face-based security solutions

Want to implement this idea in a business?

We have generated a startup concept here: FaceGuard.