Face anti-spoofing improves generalization using bonafide-centric learning
DINOFAS: Bonafide-Centric Representation Learning for Face Anti-Spoofing
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Face recognition systems can be tricked by fake images or videos, which is a security risk. This paper presents a method called DINOFAS that focuses on learning what genuine faces look like rather than just spotting known fakes. By separating real face features from attack patterns and using special techniques to handle differences in camera and environment, their system works better when encountering new types of attacks or devices. The authors tested their approach on multiple datasets showing better accuracy and adaptability than previous methods.
What this means in practice
- •For security system developers: Implement more reliable face anti-spoofing that adapts better to unknown spoofing attacks and different camera conditions.
- •For mobile app developers: Use robust facial verification methods in apps to prevent fraud through fake or altered face inputs under various conditions.
Authors
Shakeel Muhammad Ibrahim, Muhammad Sohail Ibrahim, Muhammad Shahid Jabbar, Shujaat Khan
Abstract
Face anti-spoofing (FAS) models often learn attack- and dataset-specific cues that degrade performance under unseen acquisition conditions and presentation attacks. We present DINOFAS, a DINOv3-based framework that formulates FAS as bonafide-centric representation learning. The proposed framework combines parameter-efficient DINOv3 adaptation using low-rank adaptation (LoRA), an attention style-content disentanglement (ASCD) module for reducing domain-specific appearance variations, and a dual-branch architecture that separates prototype-based liveness learning from attack-semantic supervision. The model is jointly optimized using a bonafide-centric prototype objective, an attack-semantic classification loss applied to detached features, and orthogonality regularization. DINOFAS achieves an average HTER of 9.09\% under the standard OCIM leave-one-domain-out protocol, where training is performed on three source datasets and evaluation is conducted on the held-out unseen domain. We further evaluate synthetic-to-real generalization by training exclusively on SynthASpoof and testing on four unseen real-world datasets, where DINOFAS achieves the lowest average HTER of 21.41\% and the highest average AUC of 86.14\% among the compared methods. Evaluations on OULU-NPU and SiW-Mv2 additionally demonstrate robustness to variations in acquisition conditions, devices, and unseen presentation attacks.