EEG Signal-Based Authentication: A Performance Evaluation of Feature Extraction and Classification Techniques.

In the modern digital era, passwords remain a common yet vulnerable means of user authentication, making them susceptible to many attacks. As a result, Two-Factor Authentication (2FA) emerged, offering heightened security. However, its two-step verification process can discourage widespread adoption. This study explores the potential of Electroencephalography (EEG) signals as an innovative alternative to 2FA by combining the authentication and validation processes into a single step. Our research examines different techniques to extract features from raw EEG signals, including band power extraction, statistical features, wavelet features, and Shannon Entropy, to capture these signals’ unique and intricate patterns. Additionally, we perform a comprehensive comparative analysis of discriminative classifiers such as Support Vector Machines (SVMs), k-Nearest Neighbors (k-NNs), Multilayer Perceptron’s (MLPs), Random Forest, and Gradient Boosting to determine the most effective approach for EEG signal-based authentication. We utilize a user-friendly web application that connects with cloud resources to validate our findings and provide a tangible demonstration. This application securely receives and stores EEG signals, allowing them to be evaluated against pre-trained machine-learning models. Our findings highlight the significant potential of EEG signals as a dependable, robust, and secure approach for user authentication, paving the way toward a future where passwords and complex 2FA processes can be substituted with a more convenient and reliable EEG-based authentication system.