Diffusion model improves iris attack detection without attack data
Diff-IrisPAD: Modeling Bona Fide Iris Distributions for Unsupervised Presentation Attack Detection
Computer Vision and Pattern Recognition
Summary
Iris scanners can be tricked by fake eyes or printed images, which is a security risk. The authors introduced Diff-IrisPAD, a method that learns what real iris images look like without needing examples of fake ones. It uses a special process to rebuild clean iris pictures and spots fakes when the rebuilt image doesn’t match. Tests showed it works well on many types of attacks and different iris scanners, even for attacks it hasn’t seen before.
What this means in practice
- •For security system engineers: Enhance iris recognition system security by detecting unknown presentation attacks without needing prior attack examples.
- •For authentication device manufacturers: Incorporate an unsupervised iris PAD method to improve robustness across different sensors and attack types in biometric products.$Commercial implications: Enables sale of biometric devices with stronger, scalable iris spoof detection that does not require retraining with new attack samples.
Authors
Geetanjali Sharma, Harsh Yadav, Tanmay Deshpande, Shubham Ashwani, Aditya Nigam, Raghavendra Ramachandra
Abstract
Iris recognition systems are vulnerable to presentation attacks (PAs), including textured contact lenses, print replays, and synthetic irises. Existing PAD methods rely on supervised learning with attack data, limiting generalization to unseen attacks and new sensors. We propose Diff-IrisPAD, an unsupervised framework based on Brownian Bridge Diffusion that models bona fide iris distributions without using attack samples. The method reconstructs clean iris textures from degraded inputs and detects attacks via reconstruction inconsistency using ViT scoring and pixel-level measures. Experiments on a large-scale multi-sensor dataset (125K+ images, NIR and VIS, 8 attack types) show that Diff-IrisPAD achieves an overall ACER of 30.52, significantly outperforming diffusion-based baselines while demonstrating strong generalization to unseen attacks. These results highlight the effectiveness of diffusion-based bona fide modeling for robust and scalable iris PAD.