An Evaluation of Improvement Scheme for Boundary Problem in Cancelable Biometrics Based on Block Scramble

摘要

In biometrics authentication, re-registration and revocation are very difficult because human has limited biometrics information. For this reason, cancelable biometrics has been researched. In cancelable biometrics which is based on block scramble, a template data is represented in 2-dimensional space and it will be divided into non-overlapping blocks. These position of blocks will be scrambled. However, cancelable biometrics which is based on block scramble has a block boundary problem that the scrambled position of minutiae near to block boundary are unstable. Biometrics data is noisy inherently, minutiae points near to block boundary may belong to different blocks in enrollment phase and authentication phase. Such situation induce decreasing of matching score. In order to solve this problem, we propose a scheme that uses helper data. This helper data means perturbation that will make the minutiae which is near to the block boundary far from it. By using the helper data to unstable minutiae, the minutiae will be stable and the overall matching score will be improved.