Fingerprint spoof detection method based on Decision Tree and SVM classification techniques

The human fingerprint has many capabilities and is widely used in biometric systems since it can be sampled and accessed easily and quickly. Moreover, its samples are less prone to manipulation and damage. However, there is the possibility of some problems including physical contact, contagion and fraud occurring. In this study, an approach based on image processing and machine learning techniques is proposed and its goal is to detect a fake fingerprint from a live fingerprint in identification systems with high accuracy.In the feature extraction step, statistic-based formulas are applied to image histogram and minutia features. In the next step, a symmetric local binary method is used to enhance selection criteria. In the third step, some modifications are applied to increase efficiency of the system, including hybrid classification methods, which combine support vector machine and decision tree. Finally, the decision tree is used for purposes of classification. LiveDet2015 standard database is used for analysis. Considering the results obtained from this study, the proposed system can detect a fake fingerprint from a live fingerprint with less than 0.167 percent classification error through reducing computational costs in both the feature extraction and classification steps.

Keywords:

Biometric , Fingerprint , Support Vector machine , Decision tree