Biometric Identification Systems With Both Chosen and Generated Secret Keys by Allowing Correlation

TL;DR

This paper analyzes a biometric identification system that uses both chosen and generated secret keys simultaneously, allowing correlation between them, and characterizes its fundamental limits from an information-theoretic perspective.

Contribution

It introduces a new model allowing correlation between chosen and generated secret keys and derives the capacity region for such a system.

Findings

Correlating secret keys increases the sum rate capacity.

The minimum storage rate varies with identification and secret key rates.

Privacy leakage depends only on the identification rate.

Abstract

We propose a biometric identification system where the chosen- and generated-secret keys are used simultaneously, and investigate its fundamental limits from information-theoretic perspectives. The system consists of two phases: enrollment and identification phases. In the enrollment phase, for each user, the encoder uses a secret key, which is chosen independently, and the biometric identifier to generate another secret key and a helper {data}. In the identification phase, observing the biometric sequence of the identified user, the decoder estimates index, chosen- and generated-secret keys of the identified user based on the helper {data} stored in the system database. In this study, the capacity region of such a system is characterized. In the problem setting, we allow chosen- and generated-secret keys to be correlated. As a result, by permitting the correlation of the two secret keys, the sum rate of the identification, chosen- and {generated-secret key} rates can achieve a larger value compared to the case where the keys {do not correlate}. Moreover, the minimum amount of the storage rate changes in accordance with both the identification and chosen-secret key rates, but that of the privacy-leakage rate depends only on the identification rate.