A biometric security system for detection of fake fingerprint

Abstract

Biometrics provides an alternative solution for identification which is usually carried out using PIN-codes, passwords or ID-cards. Using fingerprint recognition system as a form of biometrics identification remains one of common, quick and simple method today. But one of the major threats is that these systems can be spoofed by using a variety of methods, therefore one of major challenge is to detect where the biometric specimen is fake or real. In this project, liveness detection to verify whether the finger is real or fake is being investigated. The liveness detection method is based on the color change that a real finger exhibits when it is pressed against a hard surface. The pressure causes the veins and capillaries in the fingertip to collapse and restrict the flow of blood into that pressed region. This causes the skin region around to exhibit a change in color from reddish to whitish. Y b r C C color space was used as it provided consistent and good results for the color change. To enhance the color change, a skin-color model algorithm was developed which provided the first level of fake finger rejection. Fake fingers made from red, green and blue color material were classified as fake finger. Through the experiments, it was observed that different skin color such as like the light, medium or dark skin color does not have much effect on the modeling a person’s fingertip color. In our experiment, a total of 60 fake fingers and 60 real fingers were used, 20 fingers for each skin color type. Empirically, by using the dataset for the three different skin colors types with ages ranging between 20 to 40 years old, it demonstrated that the color change is universal and available to all the people regardless of gender, age and ethic background. It was observed that this approach gave a success rate of above 90% by using the training data available. The computational time for the experiment was 6 seconds on average on a core 2 duo notebook with 2 GB RAM.