A Pragmatic Authentication System Using Electroencephalography Signals

Abstract

EEG-based authentication is an emerging research field. In this work, a realistic authentication system using Electroencephalography signals, was developed aiming to show that brain signals contain sufficient information to be used in security systems. The dataset used was composed of 29 users on 4 different days via the cheap Neurosky Mindwave headset with a single dry electrode, and 10 users on 3 different days via Emotiv with 14 electrodes. Various techniques, features, and algorithms were examined to achieve the highest security. Experiments indicated that the system proposed can scale with respect to increasing number of users in the datasets. The system successfully handles users authenticating from multiple days not used in training the model with high accuracy. A false acceptance error (FA) of 3% was achieved, with a higher false rejection error (FR) of 48%, yielding an overall accuracy (ACC) of around 80% using the Mindwave dataset, and a FA of 0.3%, a FR of 13.93% resulting in an ACC of 92.88% using Emotiv. These results are promising for an authentication system, because the system is conservative, only allowing correct users to enter-even at the expense of multiple attempts-while successfully refusing to grant access to impersonating users.