Belief network design for biometric systems

Abstract

Bayesian belief networks represent a widely acceptable in biometric system design formalism for supporting reasoning when information is incomplete. Belief networks provide a coherent environment in which beliefs about target propositions can be re-evaluated using newly acquired evidence. This constitutes a fundamental prerequisite for decision making under uncertainty. This tutorial provides examples of applying the Bayesian decision profile in the multi-biometric system design, as well as in modeling attacks on biometric systems.