MIND-PRINT SECURITY: EEG-BASED BIOMETRIC AUTHENTICATION FOR CRITICAL INFRASTRUCTURE PROTECTION

Abstract

Critical infrastructure security requires highly developed authentication tools that shield against advanced cyber and physical attacks. Traditional approaches like passwords, access cards, and even more sophisticated things like fingerprints and iris scans are subject to theft, spoofing, or coercion. This paper examines the application of the electroencephalogram (EEG) signal as one of the biometric modalities in ultra-secure authentication in power systems, such as control rooms, substation, or SCADA systems. EEG patterns are non-visible, non-reproducible, and generated internally, making them more secure in high-risk environments. The M3CV EEG dataset was used to develop and test a cognitive-activity-based authentication scheme to generate a unique neural response during mental arithmetic and visual monitoring tasks. The signal was pre-processed using bandpass filtering, artefact elimination, and normalisation, and feature extraction was performed across the time, spectrogram, and time-frequency spectrogram dimensions. Both conventional machine learning (SVM, Random Forest, kNN) and deep models (CNN, LSTM) were evaluated. The findings reveal that deep learning-based models have a significantly higher performance than classical ones, and in particular, LSTM demonstrated an accuracy of 96 % and an Equal Error Rate (EER) of 0.035. Robustness studies supported these findings, showing that EEG authentication resists session-to-session consistency and anxiety-related variability. The results confirm the EEG biometrics' competitive claim to be a next-generation access security tool, commensurate with power protection schemes like IEC 61850 and NERC CIP. SCADA security architecture solution, incorporating EEG-based authentication, can grant resilience to any critical infrastructure against insider threats and sophisticated spoofing attacks.