Architectural Bias in Face Presentation Attack Detection: A Comparative Study of Vision Transformers and Convolutional Neural Networks

2026-06-16

AI Summary

This study addresses the significant performance disparity in existing face presentation attack detection systems across different ethnic groups, particularly their degraded accuracy on individuals with darker skin tones. The work presents the first empirical comparison of Vision Transformers and convolutional networks in terms of fairness for cross-ethnic face anti-spoofing. It demonstrates that a pretrained Vision Transformer (DeiT-S) substantially reduces performance gaps between demographic groups. On the CASIA-SURF CeFA multimodal dataset, DeiT-S achieves 97.27% accuracy and an Equal Error Rate (EER) of 0.86%, narrowing the Attack Classification Error Rate (ACER) gap between African and East Asian groups to just 0.13%. Moreover, it attains a Bona Fide Presentation Classification Error Rate (BPCER) of 2.89% on unseen Central Asian subjects—representing a 3.6× improvement in generalization over ResNet18—highlighting the critical role of model architecture in ensuring algorithmic fairness.

Abstract

Face Presentation Attack Detection (PAD) systems constitute a critical security layer in biometric authentication; however, existing approaches exhibit systematic performance disparities across demographic groups, disproportionately affecting individuals with darker skin tones. This paper presents a comparative empirical investigation of whether Vision Transformer architectures reduce demographic bias in face PAD systems relative to convolutional baselines. Experiments are conducted on the CASIA-SURF Cross-Ethnicity Face Anti-Spoofing (CeFA) dataset. Three architectures are evaluated: a Multimodal ViT-Tiny trained from scratch, a ResNet18 CNN baseline, and a pretrained DeiT-S fine-tuned on CeFA across African, East Asian, and zero-shot Central Asian demographic groups. DeiT-S achieves the highest overall accuracy of 97.27% and the lowest EER of 0.86%, outperforming ResNet18 at 90.15% accuracy. In terms of fairness, DeiT-S reduces the inter-ethnic ACER gap between African and East Asian subjects to 0.13%, compared to 0.75% reported in an LBP-based work [6], representing an 83% reduction. Most notably, while ResNet18 records a BPCER of 10.44% on zero-shot Central Asian subjects, DeiT-S maintains 2.89% on the same unseen group, demonstrating a 3.6x generalization advantage. These results suggest that pretrained Vision Transformers achieve superior PAD accuracy, produce smaller demographic performance gaps, and generalize more equitably across unseen demographic groups, indicating that cross-demographic fairness in PAD may partly be influenced by architectural design.

Problem

• Face Presentation Attack Detection
• Demographic Bias
• Fairness
• Cross-Ethnicity
• Biometric Authentication

Innovation

• Vision Transformers
• Face Presentation Attack Detection
• Demographic Bias
• Cross-Ethnic Generalization
• Architectural Fairness