Fingerprint Liveness Detection Using Directional Ridge Frequency with Machine Learning Classifiers

As security plays very important role in information era, classical security methods based on password, PIN, key, cards etc. can easily be lost, forgotten, stolen and cannot differentiate between lawful/genuine user and unlawful user. These problems can be resolved with biometric authentication. But, Biometric authentication has a threat of spoofing which, makes biometric Liveness detection very essential. It is the ability of biometric system to distinguish between spoof and real biometric sample of person. Liveness detection is one of the countermeasures needed to be taken to avoid spoof attacks to maintain integrity of biometric system and enhance biometric security. Biometric security can easily be overtaken by providing fake physical biometric to sensors. Fingerprint can easily be spoofs from some materials like silicone, glue and gelatin. To detect such spoofing different algorithms are used which broadly are classified into two approaches like hardware centric and software centric. Software centric techniques are relatively better in cost-effective ness and users have lesser interfaces (extremely vital features needed in liveness detection The paper proposes use of directional ridge frequency as features for livness detection of fingerprint with machine learning classifiers. The ridge frequencies are considered in horizontal, vertical, forward diagonal and backward diagonal directions for feature extraction. Here support vector machine(SVM), Random forest(RM), multilayer perceptron (MLP), J48 and Navies Bayes are used for classification. Experimental result have shown that random forest has given better liveness detection closely followed by SVM.