Foundation Models for Face Presentation Attack Detection: A Unified Linear-Probing Benchmark

2026-07-29

AI Summary

This work addresses the significant performance degradation of face presentation attack detection (PAD) under cross-dataset evaluation due to domain shift and limited labeled data. The authors propose a unified linear probing evaluation protocol that freezes the backbones of 24 pre-trained vision foundation models—including self-supervised ViTs, vision-language models, and supervised CNNs—and trains only a lightweight linear classifier head. This framework systematically assesses the models’ representation quality and cross-domain generalization across the four datasets in the MCIO benchmark. Their experiments, conducted for the first time under a consistent protocol, reveal the inherent PAD-relevant information embedded in foundation models as well as their domain adaptation limitations: InternViT-6B achieves the lowest average intra-domain error rate, while CLIP ViT-B/32 offers the best trade-off between cross-domain transferability and computational efficiency, underscoring the necessity of explicit adaptation to fully unlock the PAD potential of foundation models.

Abstract

Face presentation attack detection (PAD) remains challenging under cross-dataset evaluation, where domain shift degrades models trained on a single dataset. The scarcity of large-scale labeled data motivates adapting pretrained vision models rather than training task-specific architectures from scratch, raising a fundamental question: do general-purpose vision foundation models encode PAD-relevant information accessible with minimal task-specific training? To investigate, we systematically evaluate 24 frozen encoders, including self-supervised vision transformers, vision-language encoders, and supervised CNNs, using a unified linear-probing protocol on the MCIO benchmark (MSU-MFSD, CASIA-FASD, Replay-Attack, OULU-NPU). The backbone remains fixed, and only a lightweight linear head is trained to isolate the PAD information already present in the pretrained representation. %We report intra- and cross-dataset performance, along with accuracy-compute trade-offs, relative to two specialist PAD baselines. Results show that frozen foundation-model representations can support strong intra-dataset PAD performance with only a linear classifier, but this performance does not reliably transfer across datasets. Model scale is beneficial within several families, although the effect is not monotonic and is strongly mediated by architecture and pretraining. InternViT-6B achieves the lowest mean intra-dataset error, whereas CLIP ViT-B/32 offers the most favorable cross-dataset transfer-compute trade-off among the evaluated probes. These findings suggest that while pretrained representations contain PAD-relevant information, explicit adaptation remains necessary to address domain shift.

Problem

• face presentation attack detection
• cross-dataset evaluation
• domain shift
• foundation models
• linear probing

Innovation

• foundation models
• face presentation attack detection
• linear probing
• cross-dataset generalization
• frozen encoder