A Cancellable Ranking Based Hashing Method for Fingerprint Template Protection.

Zhe Jin,Jung Yeon Hwang,Soohyung Kim,Sangrae Cho,Yen-Lung Lai,Andrew Beng Jin Teoh
DOI: https://doi.org/10.1007/978-3-319-90775-8_30
2018-01-01
Abstract:Despite a variety of theoretical-sound techniques have been proposed for biometric template protection, there is rarely practical solution that guarantees non-invertibility, cancellability, non-linkability and performance simultaneously. In this paper, a cancellable ranking based hashing is proposed for fingerprint template protection. The proposed method transforms a real-valued feature vector into an index code such that the pairwise-order measure in the hashed codes are closely correlated with rank similarity measure. Such a ranking based hashing offers two major merits: (1) Resilient to noises/perturbations in numeric values; and (2) Highly nonlinear embedding based on the rank correlation statistics. The former takes care of the accuracy performance mitigating numeric noises/perturbations while the latter offers strong non-invertible transformation via nonlinear feature embedding from Euclidean to Rank space that leads to toughness in inversion yet still preserve accuracy performance. The experimental results demonstrate reasonable accuracy performance on benchmark FVC2002 and FVC2004 fingerprint databases. The analyses justify its resilience to inversion, brute force and preimage attack as well as satisfy the revocability and unlink ability criteria of cancellable biometrics.
What problem does this paper attempt to address?