High-accuracy user identification using EEG biometrics.

Area of Science:

• Neuroscience
• Biometrics
• Machine Learning

Background:

• Consumer-grade electroencephalography (EEG) devices offer a potential avenue for non-invasive biometric authentication.
• Event-related potentials (ERPs), specifically the P300 component, are known neural markers that can vary between individuals.

Purpose of the Study:

• To evaluate the efficacy of consumer-grade EEG for user identification and authentication.
• To determine the optimal machine learning approach for EEG-based biometrics.

Main Methods:

• Analysis of 14-channel EEG data from 25 subjects, focusing on the P300 component of event-related potentials (ERPs).
• Application of various machine learning techniques, including dimensionality reduction and classification algorithms.
• Comparison of identification accuracy using single and multiple ERP epochs.

Main Results:

• Statistical significance of the P300 component was confirmed in the EEG data.
• An initial user identification accuracy of 72% was achieved using a single 800 ms ERP epoch.
• Combining multiple epochs significantly improved identification accuracy to over 96.7%.

Conclusions:

• Consumer-grade EEG devices demonstrate significant potential for reliable user identification.
• Machine learning models, particularly when analyzing multiple epochs, can achieve high accuracy for EEG-based authentication.
• This technology could pave the way for novel, non-invasive security solutions.