A Multi-Filter Fingerprint Matching Framework for Cancelable Template Design
Quang Nhat Tran, Jiankun Hu
Abstract
Despite the ubiquity in the use of biometrics due to its many advantages against traditional methods such as password or token, the emerging cancelable biometric methods, which are designed to protect the biometrics are still exposed to certain threats. Attack via Record Multiplicity (ARM) is one of those. In this paper, we propose a novel framework that possesses two layers of authentication to improve the matching performance of a fingerprint authentication system in the cancelable template setting. In addition, a multi-filter fingerprint matching scheme is devised to deal more effectively with low-quality fingerprint images. Two techniques that are capable of defending against the heinous ARM are also introduced. Security analysis on the system's capability against the hill-climb attack and pre-image attack is also provided. The proposed scheme has been evaluated over public datasets FVC2002-DB1, FVC2002-DB2, FVC2002-DB3, and FVC2004-DB2. It has achieved the best result compared with the state-of-art methods. The source code for this framework is available on demand.