EEG Data for User Authentication with Multi-Class and One-Class Classifiers

L. Hernández-Álvarez, S. Caputo, L. Mucchi, and L. Hernández Encinas
Actas VII Jornadas Nacionales de Investigación en Ciberseguridad (JNIC'2022), 205-208. J. M. de Fuentes, L. González, J. C. Sancho, A. Ayerbe and M. L. Escalante (Eds.), Bilbao, Junio 27--29, 2022

Nowadays, the development of user authentication protocols is a hot topic, due to the importance of authentication mechanisms in online services as bank applications, online shop- ping or personal and professional document requests. Biometric information is commonly combined with Artificial Intelligence (Machine Learning and Deep Learning) methods to develop these systems. Nevertheless, they are usually based on Multi–Class classifiers, which need the impostor’s information in order to be trained. The access to the impostor’s information is an unrealistic assumption and, therefore, in this ongoing research we propose the construction of more realistic user authentication models using One–Class classifiers, and compare their performance with Multi–Class classifiers. Moreover, we also pretend to evaluate the contribution of different sensor locations and brain waves, and define the best model for a secure and a usable user authentication system.



This work was supported in part by the Spanish State Research Agency (AEI) of the Ministry of Science and Inno- vation (MCIN), project P2QProMeTe (PID2020-112586RB- I00/AEI/10.13039/501100011033); in part by Comunidad de Madrid (Spain) through Project CYNAMON, grant No. P2018/TCS-4566-CM, both co-funded by the European Re- gional Development Fund (ESF, FEDER and ERDF, EU); in part by the European Union’s Horizon 2020 Research and Innovation Program under Grant 872752 and under Grant 101017141. L.H.A. would like to thank CSIC Project CAS- DiM for its support.