EEG-biometric authentication system using multi-paradigm neurophysiological measurements and artificial intelligence

Electroencephalography (EEG) provides a promising physiological modality for biometric authentication due to its inherent neural uniqueness and resistance to spoofing attacks. However, reliable EEG-based authentication remains challenging because brain signals are highly non-stationary and vary across cognitive states. To address this limitation, this study proposes a multi-paradigm neurophysiological measurement protocol designed to capture stable subject-specific EEG signatures across diverse brain conditions. EEG signals were recorded from participants during six complementary paradigms, including eyes-open and eyes-closed resting states, left and right-hand motor tasks, working-memory encoding, steady-state visual evoked potentials (SSVEP), and steady-state auditory evoked potentials (SSAEP). The recorded signals were preprocessed and segmented into 2-s epochs with 50% overlap. We extracted 28 quantitative descriptors per channel from time, frequency, time-frequency, and nonlinear domains. These descriptors were used to evaluate several decision models, including Mahalanobis distance-based template matching, logistic regression (LR), support vector machine, random forest, and a one-dimensional convolutional neural network. The performance of the proposed measurement system was analyzed under both single-paradigm and multi-paradigm configurations using cross-session validation. Experimental results show that the multi-paradigm measurement protocol reduces the equal error rate (EER) by 2% compared with single-paradigm systems. The LR decision model achieved the best performance with an EER of 1.6% (95% confidence interval: 1.48−1.73), with false rejection rates of 2.06% and 9.01% at false acceptance rates of 1% and 0.1%, respectively. Feature analysis indicated that Hjorth mobility and Mel-frequency cepstral coefficients derived from fronto-central cortical channels provide the most reliable neural identity descriptors. These results demonstrate that multi-paradigm EEG acquisition improves the stability of neural measurements across cognitive states and enables reliable biometric authentication in dynamic conditions.