Secure speech biometric templates for user authentication

The security of biometrics against attacks is a serious concern in biometric personal authentication systems. In particular, the security of biometric templates is a topic of rapidly growing importance in the area of user authentication. The authors first demonstrate the security of a protected speech biometric template and devised an algorithm to attack a speech biometric user authentication system where templates are protected by a cryptographic framework. The experimental result showed an improvement for attackers in gaining access to the system. Then, a way to combine a password with a speech biometric cryptosystem is proposed. The authors present two schemes to enhance verification performance in a biometric cryptosystem using a password. Both can resist a password brute-force search if the biometrics are not compromised. Even if the biometrics are compromised, the attackers have to make many more attempts in searching for cryptographic keys in the system described in this study, compared to a traditional password-based approach. Finally, it is shown that the error rate of the proposed scheme is the same as in a traditional password-based approach even when genuine biometrics or templates are compromised.