BrainID: Development of an EEG-based biometric authentication system

Authentication is a crucial consideration when securing data or any kind of information system. Though existing approaches for authentication are user-friendly, they have vulnerabilities such as the possibility & criminally threatening a user. We propose a novel approach which uses Electroencephalogram (EEG) brain signals for an authentication process. Unique features of EEG data for distinguishing brain activities can potentially be used to authenticate a user. Compared to other biometric systems, this approach is very robust and more secure because response is significantly changed according to instantaneous mental condition. In the proposed approach, the system user is asked to visualize a number while corresponding EEG signals are captured. Captured signals are used to train the system, compared in the authentication process. This approach mainly focuses on the 8 to 30 Hz Alpha and Beta combined frequency band across all EEG channels, since it is the most appropriate band for EEG signals in Brain Computer Interfaces (BCI). Common Spatial Patterns (CSP) values were used as main features to train the model. Linear discriminant analysis (LDA) was used as a classification algorithm for a given set of user data. A trained set of models for each user is embedded in the system as a parameter database. Each user selects the profile attached to their trained model before the authentication session. Eventually, a trained model can authenticate a user after memorization of pre-determined four digits which the user is asked to think at the first stage of the process. The maximum accuracy recorded with the existing data was 96.97%.