Privacy-Preserving Indexing of Iris-Codes with Cancelable Bloom Filter-based Search Structures

Abstract

Protecting the privacy of the enrolled subjects is an important requirement expected from biometric systems. In recent years, numerous template protection schemes have been proposed, but so far none of them have been shown to be suitable for indexing (workload reduction) in the computationally expensive identification mode. This paper presents a, best to the authors' knowledge, first method in the scientific literature for indexing protected iris templates. It is based on applying random permutations to Iris-Code rows, and subsequent indexing using Bloom filters and binary search trees. In a security evaluation, the unlinkability, irreversibility and renewability of the method are demonstrated quantitatively. The biometric performance and workload reduction are assessed in an open-set identification scenario on the IITD and CASIA-Iris-Thousand datasets. The method exhibits high biometric performance and reduces the required computational workload to less than 5% of the baseline Iris-Code system.