Decoupled learning improves face attack detection with smart glasses
Decoupled Representation Learning for Face Presentation Attack Detection under Smart Glass Capture
Computer Vision and Pattern Recognition
Summary
Face recognition systems can be tricked using fake masks, and detecting these tricks gets harder when the camera is mounted on smart glasses due to movement and changing surroundings. The authors created a large new dataset of face pictures taken from smart glasses in various settings with real and fake faces. They developed a new method that separates features related to attacks from those caused by the changing environment, improving accuracy. Their tests show this method works better than others in spotting attacks while correctly accepting real faces, even when conditions vary.
What this means in practice
- •For smart device developers: Implement improved face anti-spoofing methods robust to smart glasses' changing viewpoint and motion conditions.$Commercial implications: Enables development of secure user authentication for smart glasses and wearable devices vulnerable to face mask attacks.
- •For security system engineers: Enhance camera-based face recognition systems to better reject spoof attacks under varying capture conditions like background and lighting.
Authors
Raghavendra Ramachandra, Narayan Vetrekar, Neha Naik, Rajendra Gad
Abstract
Face presentation attack detection is typically studied under fixed camera or handheld acquisition, whereas smart glass capture introduces a more challenging setting with ego motion, changing viewpoint, background clutter, distance variation, and less controlled illumination. To address this challenge, we present \emph{SmaGla-FPAD}, a smart glass based face presentation attack database acquired using Ray-Ban Meta Wayfarer Smart Glasses across indoor, outdoor, and in-car settings. The database contains 297,000 sample images with both bona fide and silicone mask based attacks. We further propose \emph{DeCoupPAD-FAS++}, a decoupled representation learning framework that separates spoof related evidence from nuisance sensitive capture variation through complementary attack and style branches, pseudo-domain supervision, adversarial suppression, and prompt guided semantic alignment. Experiments on SmaGla-FPAD show that the proposed method achieves the best balance between attack rejection and bona fide acceptance in both in domain and cross scenario evaluation. Ablation results further confirm the contribution of disentanglement, pseudo-domain supervision, and prompt based regularisation under smart glass capture.