Computationally secure authentication with noisy data

Abstract

In this chapter we discuss authentication techniques involving data such as biometrics, which are assumed to be typical (essentially unique) for a particular person (or physical object). The data are captured by a sensor or measuring device, which is an imperfect process introducing some noise. Upon enrollment of a user, a sample of the noisy data is captured and stored as a template. Later, during authentication, another sample of the noisy data is captured and matched against the stored template. © 2007 Springer-Verlag London.