Evaluation of Normalization in Fake Fingerprint Detection with Heterogeneous Sensor

Along with the development of a security system model using fingerprint authentication, a new problem arises that is the use of fake fingerprints. Sensors used to capture fingerprint images also influence recognition results. This study aims to detect fake fingerprint images, with a high degree of accuracy, and analyze the effect of normalization on two sensors from different fingerprint images. The dataset used comes from ATVS public data, where the first dataset character uses an optical sensor and a second dataset character using a thermal sensor. Min-max normalization method is proposed to improve the performance of fake fingerprint image detection, with different sensor conditions. To investigate the effect of normalization, we evaluate with the GLCM (Gray Level Co-Occurrence Matrix) image feature, in several different classification methods such as KNN (K-Nearest Neighbor), SVM (Support Vector Machine), and NN (Neural Network). Increased accuracy results obtained in this study with normalization up to 24%.