Key Generation in a Voice Based Template Free Biometric Security System

Abstract

Biometric systems possess major drawbacks in their inability to be revoked and re-issued as would be the case with passwords if lost or stolen. The implication is that once a biometric source has been compromised, the owner of the biometric as well as the data protected by the biometric is compromised for life. These concerns have necessitated research in template free biometrics, which exploits the possibility of directly encrypting the biometric data provided by the individual and therefore eliminates the need for storing templates used for data validation, and thus increasing the security of the system. Template free system function in stages- (a) Calibration during which feature distribution maps of typical users are generated from known biometric samples without storing any personal information, and (b) Operation which uses the feature distribution maps as a reference to generate encryption keys from samples of previously unseen users and also rebuilds the key when needed from new sets of previously unseen samples. In this report, we used a combination of stable features from the human voice to directly generate biometric keys using a novel method of feature concatenation to combine the binary information at the operation phase. The stability of the keys is superior to current key generation methods and the elimination of biometric templates improves the safety of personal information and in turn increases users’ confidence in the system.