On the Vulnerability of an EEG-based Biometric System to Hill-climbing Attacks Algorithms' Comparison and Possible Countermeasures

In this paper we analyze the vulnerability to hill-climbing attacks of a biometric recognition system based on electroencephalography (EEG). It is assumed that an attacker can access the scores produced by the employed matcher, and use them to control the generation of synthetic EEG templates until achieving a successful authentication. TO this aim, different general approaches relying on function optimization are evaluated and compared in terms of authentication success rate and average number of required attempts. The possibility of increasing the system robustness against this kind of attacks, without significantly affecting its recognition performance, is also investigated.