Cancellable Fingerprint Template Generation Using Rectangle-Based Adjoining Minutiae Pairs

Abstract

Cancellable fingerprint templates effectively protect original fingerprint data by revoking an accorded template and reissuing a new template. Alignment-free cancellable templates require no image pre-alignment and therefore does not go through from inaccurate singular point detection. In our proposed method, we focused on generating a cancellable template which is alignment-free. The template is generated by the building of R rectangles by varying the directions over every minutia succeeded by the computation of translation invariant and rotation invariant adjoining relation. The computed feature set is quantized & mapped into a cube to produce a binary string. Further, we apply modulo operation on the generated bit string to get reduced bit string which mitigates the risk of the ARM (Attack via Record Multiplicity). Later, we apply Discrete Fourier Transform (DFT) to convert reduced binary string into a complex vector. The result is then multiplied by an arbitrary matrix to produce the cancellable template. We evaluated proposed scheme on databases FVC 2004 DB1-DB3 & FVC 2002 DB1-DB3 and results fulfills the conditions of Biometric Template Protection Scheme(BTPS) and it gives competitive performance(in terms of EER) when compared to existing methods.