🤖 AI Summary
Password-based authentication in XR environments disrupts immersion and encourages weak password usage. To address this, we propose an unobtrusive multimodal biometric authentication system specifically designed for XR, the first to jointly leverage consumer-grade eye-tracking (Tobii) and electroencephalography (EEG) (OpenBCI) signals. Our method employs a dual-branch siamese neural network architecture to enable cross-modal feature complementarity and end-to-end matching; synchronized signal acquisition and joint modeling mitigate noise from low-fidelity hardware. Evaluated on a 30-subject dataset, the system achieves an equal error rate (EER) of 0.298%, representing an 83.6% reduction over eye-tracking alone and a 93.9% reduction over EEG alone. This demonstrates substantial improvements in both security and user experience consistency—eliminating explicit authentication steps while preserving XR immersion.
📝 Abstract
The increasing adoption of Extended Reality (XR) in various applications underscores the need for secure and user-friendly authentication methods. However, existing methods can disrupt the immersive experience in XR settings, or suffer from higher false acceptance rates. In this paper, we introduce a multimodal biometric authentication system that combines eye movement and brainwave patterns, as captured by consumer-grade low-fidelity sensors. Our multimodal authentication exploits the non-invasive and hands-free properties of eye movement and brainwaves to provide a seamless XR user experience and enhanced security as well. Using synchronized eye and brainwave data collected from 30 participants through consumer-grade devices, we investigated whether twin neural networks can utilize these biometrics for identity verification. Our multimodal authentication system yields an excellent Equal Error Rate (EER) of 0.298%, which means an 83.6% reduction in EER compared to the single eye movement modality or a 93.9% reduction in EER compared to the single brainwave modality.