FoundPAD: Foundation Models Reloaded for Face Presentation Attack Detection

2025-01-06

AI Summary

To address the weak cross-scenario generalization and heavy reliance on labeled data in facial presentation attack detection (PAD), this paper pioneers the integration of foundation models (FMs) into PAD. We propose a LoRA-augmented classification head joint-training architecture, enabling efficient low-rank adaptation and end-to-end optimization of the FM—facilitating training even with extremely limited real-world data or purely synthetic data. Our approach achieves state-of-the-art generalization performance across multiple unseen domains, significantly reducing dependence on large-scale annotated datasets. The code is publicly released to ensure full reproducibility. Key contributions include: (1) a paradigm shift in PAD driven by foundation models; (2) a novel LoRA-classification head co-optimization mechanism; and (3) a lightweight, synthesis-data-enabled training framework that delivers high generalization with minimal real-data supervision.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Although face recognition systems have seen a massive performance enhancement in recent years, they are still targeted by threats such as presentation attacks, leading to the need for generalizable presentation attack detection (PAD) algorithms. Current PAD solutions suffer from two main problems: low generalization to unknown cenarios and large training data requirements. Foundation models (FM) are pre-trained on extensive datasets, achieving remarkable results when generalizing to unseen domains and allowing for efficient task-specific adaption even when little training data are available. In this work, we recognize the potential of FMs to address common PAD problems and tackle the PAD task with an adapted FM for the first time. The FM under consideration is adapted with LoRA weights while simultaneously training a classification header. The resultant architecture, FoundPAD, is highly generalizable to unseen domains, achieving competitive results in several settings under different data availability scenarios and even when using synthetic training data. To encourage reproducibility and facilitate further research in PAD, we publicly release the implementation of FoundPAD at https://github.com/gurayozgur/FoundPAD .

Problem

• Presentation Attack Detection
• Facial Recognition Security
• Algorithm Performance Degradation

Innovation

• FoundPAD
• LoRA weight tuning
• Simultaneous classification learning