Brain signals have recently been proposed as a strong biometric due to their characteristics such as, uniqueness, permanence, universality, and confidentiality. There are many factors that affect the stability of EEG
signals as a biometric for example, using different recording devices, variation in participant emotional
states, performing different mental tasks and recording in temporally spaced sessions. Due to the nonstationary nature of EEG signals, there are still speculations about the stability of using them for generating
a unique and repeatable cryptographic key. The challenge that faces all biometric based crypto-systems is to
overcome the variation in biometric itself over time and to generate multiple unique keys from the same
observation. In this work, we investigate the stability of using EEG signals as a biometric for both personal
authentication and cryptographic key generation. The authentication process was tested using three datasets AMIGOS, DEAP, and SEED. Achieving accuracy of 96:23%; 98:85%, and 99:89% respectively. The key
generation process generates a set of different keys with different lengths from the same observation.
Each key is unique and repeatable. The generated keys were examined using NIST test suite, scale index test,
and autocorrelation test. Time complexity analysis of the key generation process was performed.