Evaluation of Fingerprint Liveness Detection by Machine Learning Approach - A Systematic View

Abstract

Recently, fake fingerprint detection is a challenging task in the cyber-crime sector in any developed country. Biometric authentication is growing in many sectors such as internet banking, secret file locker, etc. There spoof fingerprint detection is an essential element that is used to detect spot-on fingerprint analysis. This article focuses on the implementation and evaluation of suitable machine learning algorithms to detect fingerprint liveness. It also includes the comparative study between Ridge-let Transform (RT) and the Machine Learning (ML) approach. This article emphasis on research and analysis of the detection of the liveness spoof fingerprint and identifies the problems in different techniques and solutions. The support vector machine (SVM) classifiers work with indiscriminate loads and confined grayscale array values. This leads to a liveness report of fingerprints for detection purposes. The SVM methodology classifies the fingerprint images among more than 50K of real and spoof fingerprint image collections based on this logic. Our proposed method achieves an overall high accuracy of detection of liveness fingerprint analysis. The ensemble classifier approach model is proving an overall efficiency rate of 90.34 % accurately classifies samples than the image recognition method with RT. This recommended method demonstrates the decrement of 2.5% error rate when compared with existing methods. The augmentation of the dataset is used to improve the accuracy to detect. Besides, it gives fake fingerprint recognition and makes available future direction.