Machine Learning and Deep Learning Techniques for Distinguishing Between Genuine Users and Impostors in EEG-Based Authentication Using Pattern Lock

Traditional validation methods are vulnerable to different security risks, rendering them less dependable in protecting confidential information. EEG-based biometric systems offer a potential solution using unique neural patterns for user authentication but require effective analytical techniques to optimize performance and security. This investigation examines how well various machine learning models, including Neural Networks of Deep Learning, perform in EEG-based authentication systems, emphasizing their ability to distinguish genuine users from impostors. The research utilized a dataset comprising EEG data from controlled experiments with predefined authentication tasks. The tested models included SVM, Logistic Regression, NN, Naïve Bayes, KNN, and Decision Trees, focusing on optimizing accuracy, precision, and recall. The experiments showed significant variability in model performance. The KNN model attained the maximum accuracy of 100%, whereas SVM, Logistic Regression, and Naïve Bayes exhibited a lower level of performance at approximately 33.2%, 46.7<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, and 40.3%. Decision Trees and Neural Networks showcased notable effectiveness, achieving 99.4% and 97% accuracy, respectively. These findings highlight the promise of integrating various models to bolster the resilience of EEG-based authentication systems. The findings suggest that while some models like KNN and Decision Trees perform exceptionally well, integrating multiple machine-learning techniques could provide a more balanced and secure authentication system. This approach not only leverages the strengths of various models but also addresses their limitations, leading to improved security and reliability in EEG-based biometric authentication.