Cancelable fingerprint templates with Delaunay triangle-based local structures

Abstract

Security of biometric templates in a fingerprint authentication system is highly critical because the number of fingers of a person is limited and the raw fingerprints cannot be reset or replaced. A cancelable template is an efficient and powerful means to provide template protection. However, in most cancelable fingerprint templates, the many-to-one based non-invertible transformation acts on each single minutia, which may greatly decrease the discriminative capability of its feature representation, and more importantly, single minutia is more sensitive than a local structure to the non-linear distortion present in a fingerprint image. In this paper, we propose a cancelable template by utilizing some nice properties of Delaunay triangle-based local structures, e.g., excellent local structural stability, to achieve satisfactory performance. The non-invertible transformation is applied to each Delaunay triangle-based local structure rather than to each individual minutia so as to mitigate the influence of non-linear distortion and retain the discriminative power of the feature data after the many-to-one mapping. Experimental results on the publicly available databases demonstrate the validity of the proposed scheme. © Springer International Publishing Switzerland 2013.