NeuroShield: A Device-Agnostic Foundation Model for EEG Authentication

NeuroShield is a device-agnostic foundation model for EEG authentication, designed to overcome the limitations of existing models tied to specific acquisition settings. Utilizing a dual-stage transformer architecture, it was pretrained on 15,762 subjects and 28,116 sessions from three public EEG datasets, achieving a reduction in equal error rate by 0.44–8.06 percentage points compared to state-of-the-art methods after fine-tuning. This model's ability to generalize across variable-channel layouts and longer segments enhances its reusability and adaptability, making it a significant advancement for practitioners in EEG-based identity verification.