Binding Cryptographic Keys into Biometric Data: Optimization

Abstract

Cryptography and biometry are important components of contemporary access control systems. Cryptographic systems themselves are highly reliable but they require the exact reproduction of access keys; this cannot be done by humans, while the corresponding devices might be lost or stolen. Biometric data are always with the person; however, they vary: it is impossible to obtain the same feature values. In this paper, a way is proposed to link the cryptographic key and the biometric features of the iris. This yields a two-component key such that no original component can be extracted until the biometric features close to the original ones, i.e., the data of the same person, are presented. The connecting method (coder) and the extracting method (decoder) consist of several separate steps executed successively. To select the parameters, we solve the following discrete optimization problem: under the given threshold of the false accept rate, we minimize the value of the false reject rate. The restrictions of this optimization problem are the minimal size of the coded key and the maximal size of the final key. Numerical experiments are conducted on open-access databases (DBs).