Harnessing Chain-of-Thought Reasoning in Multimodal Large Language Models for Face Anti-Spoofing

TL;DR

This paper introduces FaceCoT, a large-scale multimodal dataset with chain-of-thought annotations for face anti-spoofing, and proposes a novel training strategy that significantly improves model robustness and interpretability.

Contribution

It presents FaceCoT, the first comprehensive VQA dataset for FAS, and a CoT-Enhanced Progressive Learning method to leverage multimodal reasoning for better anti-spoofing performance.

Findings

Models trained with FaceCoT and CEPL outperform existing methods.

The dataset covers 14 spoofing attack types with high-quality annotations.

Experimental results show improved robustness and interpretability in FAS models.

Abstract

Face Anti-Spoofing (FAS) typically depends on a single visual modality when defending against presentation attacks such as print attacks, screen replays, and 3D masks, resulting in limited generalization across devices, environments, and attack types. Meanwhile, Multimodal Large Language Models (MLLMs) have recently achieved breakthroughs in image-text understanding and semantic reasoning, suggesting that integrating visual and linguistic co-inference into FAS can substantially improve both robustness and interpretability. However, the lack of a high-quality vision-language multimodal dataset has been a critical bottleneck. To address this, we introduce FaceCoT (Face Chain-of-Thought), the first large-scale Visual Question Answering (VQA) dataset tailored for FAS. FaceCoT covers 14 spoofing attack types and enriches model learning with high-quality CoT VQA annotations. Meanwhile, we develop a caption model refined via reinforcement learning to expand the dataset and enhance annotation quality. Furthermore, we introduce a CoT-Enhanced Progressive Learning (CEPL) strategy to better leverage the CoT data and boost model performance on FAS tasks. Extensive experiments demonstrate that models trained with FaceCoT and CEPL outperform state-of-the-art methods on multiple benchmark datasets.