FoundPAD: Foundation Models Reloaded for Face Presentation Attack Detection
Guray Ozgur, Eduarda Caldeira, Tahar Chettaoui, Fadi Boutros, Raghavendra Ramachandra, Naser Damer
TL;DR
The paper tackles cross-domain face presentation attack detection under low data regimes by leveraging foundation models. It introduces FoundPAD, a CLIP-based framework fine-tuned with rsLoRA that preserves pretrained knowledge while aligning the embedding space to PAD through a lightweight classifier. Extensive cross-dataset evaluations show FoundPAD consistently outperforming baseline strategies that train from scratch or use frozen embeddings, achieving competitive state-of-the-art results and validating performance with synthetic training data. The work demonstrates that foundation models, when carefully adapted with efficient fine-tuning methods, offer robust, data-efficient PAD with practical reproducibility.
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 .