Secured biometric verification based on fingerprint topology binary representation

The paper deals with combination of cryptographic constructions and fingerprint identification. A technique of repeatable binary string extraction from fingerprint images is suggested. The binary features from topological relations between minutiae points are extracted. For an arbitrary minutiae point, the neighboring ridges are traced until the event is encountered: minutiae or projection of minutiae. Then, these events are encoded. Thus, 50–100-bit descriptions for each minutiae point are obtained. In order to extract longer binary string, two techniques are suggested. The first one is the self-aligned technique, while the second one requires public helper. Thus, 384–756-bit binary string is extracted. The strings have approximately 20% erroneous bits, which are corrected using two-layer Bose–Chaudhuri–Hocquenghem (BCH) major voting codes. The experiments on FVC2002 DB1 dataset show that 20–40-bit error-free binary string can be reproduced from genuine fingerprint with 90 percent success rate.