Introducing Nylon Face Mask Attacks: A Dataset for Evaluating Generalised Face Presentation Attack Detection

The introduction of the Nylon Face Mask Attacks dataset marks a significant advancement in the evaluation of face recognition systems, which are increasingly used for security and authentication. With 3,760 bona fide samples and 51,281 attack samples collected using an iPhone 11 Pro, this dataset simulates realistic presentation attacks using nylon face masks that closely resemble human faces. The study reveals considerable performance variability among five state-of-the-art presentation attack detection methods, emphasizing the challenges posed by these advanced spoofing techniques. As face recognition technology becomes more prevalent, understanding and mitigating its vulnerabilities is crucial. This dataset not only provides a benchmark for evaluating existing detection methods but also highlights the urgent need for the development of more robust techniques to effectively counteract emerging threats in the field of biometric security.