Machine Learning Meets EEG: A Novel Approach to PGA-Based Authentication Systems

The development of secure and effective identification/authentication systems is an ongoing challenge in the field of computer science, as new techniques and methods continue to emerge. In the context of this ever-evolving landscape, we propose a new multi-factor authentication system. The proposed...

Full description

Saved in:
Bibliographic Details
Published in:Proceedings (IEEE International Conference on Emerging Technologies and Factory Automation) pp. 924 - 929
Main Authors: Katoikos, Ioannis, Fidas, Christos A., Koukopoulos, Dimitrios
Format: Conference Proceeding
Language:English
Published: IEEE 18.12.2024
Subjects:
Authentication
Brain modeling
Electroencephalography
Electronics packaging
Feature extraction
Graphical/visual passwords
Human-centered computiug→ Laboratory experiments
Machine learning
Machine learning algorithms
Multi-factor authentication
Recording
Security
security and privacy → Biometrics
Visualization
ISSN:1946-0759
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The development of secure and effective identification/authentication systems is an ongoing challenge in the field of computer science, as new techniques and methods continue to emerge. In the context of this ever-evolving landscape, we propose a new multi-factor authentication system. The proposed system combines Picture Gesture Authentication (PGA) with simultaneous acquisition of users' brain signals (EEG) to enhance security. To assess the efficacy of the proposed model, we collected data from 14 individuals, each of whom participated in 20 trials to create a single-person authentication system. Features were extracted and multiple machine learning algorithms were trained and tested in three different scenarios. The first scenario utilized the created dataset to find the most effective machine learning algorithms and features; the second applied augmentation techniques for a more balanced dataset; the third used only the most important features. The results were encouraging, indicating that the proposed model works well in enhancing the level of security.
ISSN:1946-0759
DOI:10.1109/ICMLA61862.2024.00135