A Modified Fuzzy Fingerprint Vault Based on Pair-Polar Minutiae Structures

摘要

Biometrics-based cryptosystems are helpful tools facilitating our daily life. Examples are that people use their biometrics to identify themselves in clouds or protect passwords as well as other privacy information in mobile phones connected to Internet-of-Things. Among various biometrics, fingerprints are one used most widely. A promising type of secure cryptosystems based on fingerprints is the so called fuzzy fingerprint vaults which can bind a secret key to one fingerprint template and release the same key later with the help of a similar but not necessarily fully identical fingerprint template. However, some aspects such as the security and matching accuracy of existing fingerprint-based fuzzy vaults are still not satisfactory due to troublesome issues such as the alignment of fingerprints and minutiae quantization. In this paper we propose a modified fuzzy fingerprint vault based on pair-polar minutiae structures, aiming at achieving better performance. Recall that Li et al. (2016) presented a fuzzy vault based on pair-polar minutiae structures with two-level secure sketch in recovering the secret key. Unlike their scheme, we compress the two-level structure into one level in which minutia descriptors are exploited to match a vault and thus lower the time complexity of decoding. Our experiment evaluation shows that the proposed fuzzy fingerprint vault achieves high matching accuracy. Importantly, our vault is of strong security against existing attacks such as the brute-force attack, false-accept attack and correlation attack. Also this vault adopts the polynomial representation introduced by Dodis and Reyzin (2006) which is thus able to reduce the storage complexity.