Using electroencephalogram (EEG) signals for person authentication purpose so that people can access a security system by just thinking a pass-thought instead of typing a password is an interesting research topic. However, many factors can impact on the stability of an EEG-based person authentication system, as they may create instability in the pass-thought, yet the issue has not been comprehensively researched. In this paper, we focus on that gap by investigating the performance variations of an EEG-based person authentication system when users are in different brain states, caused by having different emotional states and users’ different experiences toward stimuli while performing mental tasks for pass-thought. Also, we speculate on whether human characteristics such as gender and age have an impact on the performance of EBPA system while users are in different brain states. The experimental results revealed that user changing emotions when logging into the system differ with that when they enroll have a negative impact on the performance of a security system. Further, the young and female groups always give higher accuracy compared to older and male groups regardless of brain state. The results encourage careful consideration of different brain states in order to build a higher security and more stable person authentication system for real world applications.