On the Stability of EEG-based Person Authentication Systems in Different Brain States

Abstract

Using EEG signals for person authentication purposes provides distinctive benefits of a new type of
biometrics that requires the users to be alive when recording, and that are unintrusive, impossible
to observe or intercept and impossible to mimic. However, EEG signals are known to be sensitive
to many factors such as user's different emotional states, user's experience toward stimuli, human
characteristics, health condition and stimulant consumption. In this research, we focus on the stability
of EEG-based person authentication systems when users are in different brain states, which
are caused by the aforementioned factors.
First, a relatively comprehensive investigation on the stability of EEG-based person authentication
systems is addressed when users are in different brain states, which are caused by mental states
with two common instances, namely emotional states and experience toward stimuli.
Secondly, the research further investigates how stable an EEG-based person authentication system
can produce "EEG passwords" repeatedly when users have different health conditions, different
stimulant consumption and human characteristics. Based on that, the research finds out which
EEG patterns related to particular brain states are better for an EEG-based person authentication
system. Also, it introduces some ideas based on the experimental results, for designing a real-world
EEG-based person authentication system that is not only stable and highly secure but also comfortable
for the users.
Evaluation experiments performed on the benchmark datasets such as DEAP, Australian EEG
and Alcoholism show that the proposed method is promising.

Date of Award	2021
Original language	English
Supervisor	Shuangzhe LIU (Supervisor) & Dat TRAN (Supervisor)