Preventing the cross-matching attack in bloom filter-based
Christian Rathgeb, Johannes Wagner, Benjamin Tams, Christoph Busch · 2015
Deployments of biometric technologies are already widely disseminated, i.e. the protection of biometric reference data becomes vital in order to safeguard individuals’ privacy. Bio- metric template protection techniques are designed to protect biometric templates in an irreversible and unlinkable man- ner (ISO/IEC IS 24745). In addition, these schemes are re- quired to maintain key system properties, e.g. biometric per- formance or authentication speed. Recently, template protec- tion schemes based on Bloom filters have been introduced and applied to various biometric characteristics, such as iris or face. While a Bloom filter-based representation of biometric templates is irreversible the originally proposed system has been exposed to be vulnerable to cross-matching attacks. In this paper we address this issue and demonstrate that any kind of Bloom filter-based representation of biometric templates can be transformed to an unordered set of integer values which enables a locking of irreversible templates in a fuzzy vault scheme from Dodis et al. which can be se- cured against known cross-matching attacks. In addition, ex- periments which are carried out on a publicly available iris database, show that the proposed scheme retains the biomet- ric performance of the original system.