Fingerprint Authenticity Classification Algorithm based-on Distance of Minutiae using Convolutional Neural Network

Hariyanto Hariyanto (Doctoral Program in Information Technology, Gunadarma University STMIK Jakarta STI&K)
Sarifuddin Madenda (Doctoral Program in Information Technology, Gunadarma University)
Sunny Arief Sudiro (STMIK Jakarta STI&K)
Tubagus Maulana Kusuma (Doctoral Program in Information Technology, Gunadarma University)

InComTech: Jurnal Telekomunikasi dan Komputer
Vol 11, No 3 (2021)

Fingerprint identification systems are vulnerable to attempted authentication fraud by creating fake fingerprints that mimic the live. This paper proposes method to detect whether a fingerprint is live fingerprint or fake fingerprint using Convolutional Neural Network (CNN). We construct a features database of distances among minutiaes of fingerprints, where the distance calculation is based-on Euclidean Distance. Furthermore, the distance features database that has been constructed is classified using the CNN. CNN is a deep learning method designed for machine learning processes so that computers recognize objects in an image and this method has capability classifying an object. The numerical results have shown that the best accuracy achieves 99.38% when the learning rate is 0.001 with the epoch of 100.