Selecting distinctive features to improve performances of multidimensional fuzzy vault scheme
CCBR'11 Proceedings of the 6th Chinese conference on Biometric recognition
| Hi-index | 0.00 |
Biometric cryptosystem has emerged as a promising solution in information security. Fuzzy vault is a widely accepted scheme that binding biometric data and cryptography effectively. In vault encoding, random chaff points which are not on the polynomial are added to protect genuine points. In vault decoding, query biometric data is generated to retrieve the genuine points by measuring the distance between the registered data and query data. However, because of the within-class differences and randomness of chaff points, it is difficult to distinguish the genuine data from chaff data. This paper proposes a novel 3D fuzzy vault scheme that selects unused data from template feature vector and inserts it into the genuine points. In our scheme, every point in the vault has three dimensions, two of which could be used for distance measurement. The precision of the data matching is improved by the new inserted data and genuine points could be recognized more accurately. Experimental results based on HA-BJTU database show the better performances compared to traditional fuzzy vault.