Assessing Authentication Performance in EEG Recordings: An Information Theory Metrics Approach Across Experimental Conditions

The growing integration of physiological signals into Healthcare IoT is driven by the remarkable expansion of the wearable technology market. However, ensuring the security of wireless communication in these systems, especially for applications involving sensitive health information, remains a critical concern. This study delves into the use of information theory metrics, specifically mutual information, to explore the potential of electroencephalogram (EEG) signals as biometric identifiers for individual authentication in diverse experimental conditions. Our analysis employs entropy measurements to assess the information content of EEG features extracted from a PhysioNet database dataset, including resting-state EEG signals and auditory stimuli experiments. Moreover, the use of mutual information helps quantifying the statistical dependence between EEG features and individual identity, revealing that auditory stimulation generally enhances authentication performance compared to non-stimulated conditions. Additionally, employing inear headphones for auditory stimuli delivery demonstrates a marginal improvement. This paper presents a simplified yet effective approach to evaluate EEG signal performance for individual authentication, contributing valuable insights to the broader understanding of physiological signal authentication in diverse experimental settings.