Sammendrag
As automated fingerprint recognition systems gain popularity, the proliferation of information about unchangeable biometric characteristics causes serious privacy and security concerns. This information may enable an impostor to create a matching fingerprint, and the stored information should therefore be considered extremely sensitive.
This thesis explores a novel method for generating cancellable fingerprint templates that will impede the reproduction of a fingerprint from the stored template, and at the same time allow the same fingerprint to be reused in the case of a compromise.
During enrollment, the proposed method aligns the minutiae points of a fingerprint to a reference coordinate system using the core and principal direction, and creates a hash value based on the set of minutiae points. It then generates Reed-Solomon error correction codes which enable the reproduction of
the full set of minutiae points if a certain number of minutiae points are known. It then performs an irreversible Cartesian block transformation on the
minutiae points.
During the matching process, the minutiae points of the candidate print are similarly aligned, and transformed using the same Cartesian transformation. A
standard matching algorithm is performed on the minutiae sets in the transformed space, which allows the Cartesian transformation to be reversed for the matching minutiae points in the enrolled template. Using the Reed-Solomon error-correction codes generated during enrollment, the entire minutiae point set of the enrolled print can be recreated, provided enough minutiae points could be correctly reversed.
Thus, a matching candidate fingerprint allows an otherwise irreversible transformation on the enrolled print to be reversed. The same hash value created for the fingerprint during enrollment can thus be re-generated when a matching fingerprint is presented.
A proof-of-concept implementation of the method is presented and tested. Although the recognition accuracy of the proposed method was found to be inferior to comparable traditional fingerprint recognition methods, the method nonetheless