Date: 2008/3/11(Tue) 15$B!'(B30$B!A(B17$B!'(B00
Place: Collaboration Room #7 (Information Science Building, 5th floor)
Name: Akira Otsuka
National Institute of Advanced Industrial Science and Technology(AIST).
Title: Biometrics Secuity - Wolf Attack Probability -
Abstract: A biometric authentication system automatically authenticates an
individual by using human's physiological or behavioral characteristics.
Fingerprint, face, iris, and vascular patterns are the typical physiological characteristics. Recently, the use of the biometric authentication systems has spread in various services such as the access control for a mobile phone and PC, the customer's authentication at a bank's ATM (automated tellers machine) terminal, the immigration control at an airport, and so on. Unfortunately, a security evaluation method for a biometric authentication system has not been sufficiently established so far. In order to realize a reliable biometric authentication system with high security, it is necessary to develop the security evaluation method as the priority issue. Our main research objective is to develop such a rigourous security method and, on the basis of our method, propose a new biometric authentication technique that is verifiable by the third party. Especially, we are now focusing on how to develop a quantative security measure against impersonation attacks. As a recent result, we have proposed a ``wolf attack probability (WAP)" as a new security measure. WAP is the maximum success probability of a wolf attack, which is an attempt to exhastively impersonate a victim using wolves (i.e., exhastive attack with input samples that shows high similarity to most of templates). Moreover, we have applied the wolf attack to a finger-vein-pattern matching algorithm, a typical fingerprint-minutiae matching algorithm, and iris-recognition systems and showed that the success probability of impersonation is surprisingly higher than the previously believed impersonation probability called FAR (False Acceptanc Ratio).