Design, implementation & evaluation of cancellable EEG templates using biohashing methods
Loading...
Date
2026-02-09
Authors
Γαλάνης, Ανδρέας
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis outlines the design, implementation, and evaluation of a biometric authentication system based on EEG that prioritizes privacy through the use of cancellable templates. The proposed framework seeks to improve security and revocability in EEG biometrics, all the while ensuring that identification accuracy remains high. This research draws upon the EEG Motor Movement/Imagery Dataset, which is accessible to the public through PhysioNet. The dataset includes recordings from 109 individuals, featuring 64-channel EEG data gathered during standardized tasks involving both motor execution and imagery. The EEG signals underwent a series of preprocessing steps, including filtering, artifact removal, and Independent Component Analysis (ICA). This was succeeded by segmentation and feature extraction to effectively extract reliable frequency-domain features. A detailed pipeline was created in Python, utilizing the MNE library alongside custom scripts to produce straightforward templates and implement cancellable transformations. The cancellable design utilizes biohashing techniques, which involve permutation, random projection, and quantization operations that are governed by user-specific keys. This approach guarantees non-invertibility, allows for revocation, and promotes diversity among the stored templates. The assessment utilized established biometric metrics, including the False Acceptance Rate (FAR), False Rejection Rate (FRR), Equal Error Rate (EER), and Area Under the Curve (AUC). Additionally, it incorporated more sophisticated analyses, such as revocability checks and studies on impostor correlations. The experimental findings indicated that the cancellable templates maintained a level of accuracy that was comparable to that of their plain counterparts, while also demonstrating significant resilience against key-rotation and brute-force attacks. This research presents a structured and replicable framework for EEG-based cancellable biometrics, effectively connecting the fields of neuroscience, machine learning, and information security. The results underscore the practicality of implementing brainwave-driven authentication systems in everyday settings, all while safeguarding user privacy and ensuring data protection.
Description
Keywords
EEG biometrics, Cancellable templates, Biohashing, Template protection, Revocability, Electroencephalogram (EEG)