Using a Bayesian averaging model for estimating the reliability of decisions in multimodal biometrics

Abstract

The issue of reliable authentication is of increasing importance in modern society. Corporations, businesses and individuals often wish to restrict access to logical or physical resources to those with relevant privileges. A popular method for authentication is the use of biometric data, but the uncertainty that arises due to the lack of uniqueness in biometrics has lead there to be a great deal of effort invested into multimodal biometrics. These multimodal biometric systems can give rise to large, distributed data sets that are used to decide the authenticity of a user. Bayesian model averaging (BMA) methodology has been used to allow experts to evaluate the reliability of decisions made in data mining applications. The use of decision tree (DT) models within the BMA methodology gives experts additional information on how decisions are made. In this paper we discuss how DT models within the BMA methodology can be used for authentication in multimodal biometric systems.