Benchmarking Vision Foundation Models for Domain-Generalizable Face Anti-Spoofing

TL;DR

This paper benchmarks 15 pre-trained vision models for domain-generalizable face anti-spoofing, demonstrating that self-supervised vision transformers, combined with data augmentation techniques, achieve state-of-the-art results efficiently.

Contribution

It provides a systematic evaluation of vision-only foundation models for FAS, establishing a robust and efficient baseline that surpasses existing methods in cross-domain scenarios.

Findings

Self-supervised vision models like DINOv2 excel in suppressing attention artifacts.

Combined with data augmentation, the vision-only baseline achieves state-of-the-art performance.

The approach outperforms existing methods under data-constrained protocols.

Abstract

Face Anti-Spoofing (FAS) remains challenging due to the requirement for robust domain generalization across unseen environments. While recent trends leverage Vision-Language Models (VLMs) for semantic supervision, these multimodal approaches often demand prohibitive computational resources and exhibit high inference latency. Furthermore, their efficacy is inherently limited by the quality of the underlying visual features. This paper revisits the potential of vision-only foundation models to establish a highly efficient and robust baseline for FAS. We conduct a systematic benchmarking of 15 pre-trained models, such as supervised CNNs, supervised ViTs, and self-supervised ViTs, under severe cross-domain scenarios including the MICO and Limited Source Domains (LSD) protocols. Our comprehensive analysis reveals that self-supervised vision models, particularly DINOv2 with Registers, significantly suppress attention artifacts and capture critical, fine-grained spoofing cues. Combined with Face Anti-Spoofing Data Augmentation (FAS-Aug), Patch-wise Data Augmentation (PDA) and Attention-weighted Patch Loss (APL), our proposed vision-only baseline achieves state-of-the-art performance in the MICO protocol. This baseline outperforms existing methods under the data-constrained LSD protocol while maintaining superior computational efficiency. This work provides a definitive vision-only baseline for FAS, demonstrating that optimized self-supervised vision transformers can serve as a backbone for both vision-only and future multimodal FAS systems. The project page is available at: https://gsisaoki.github.io/FAS-VFMbenchmark-CVPRW2026/ .