Helper data schemes (HDS) are used in biometric authentication to extract a noise-free authentication string. One objective is to avoid that the stored template leaks little or no information about the users biometrics. A zero leakage (ZL) fuzzy extractor (FE) is often regarded as the one of the most secure solutions. A concern is the limited verication performance, e.g. a large false rejection rate (FRR). This paper studies whether systems with a non-uniform secret or with non-zero can be more attractive. We argue that the secrecy rate, i.e., secret entropy minus leakage, versus FRR is an appropriate performance trade o. By modeling the biometric verication system as a communication channel, we show that a fuzzy extractor (FE) does not achieve capacity.
|Title of host publication
|34th WIC Symposium on Information Theory in the Benelux / Third joint WIC/IEEESP Symposium on information Theory and signal processing in het Benelux, May 30-31 2013, Leuven, Belgium
|Published - 2013