Implementasi Autentikasi Biometrik Sidik Jari Menggunakan Minutiae-Based Matching Dan Algoritma Bozorth3 Scoring Untuk Keamanan Sistem Tanda Tangan Elektronik

Abstract

Electronic Signature (E-Signature) systems are increasingly used in digital administration, government services, and electronic transactions, yet most still rely on username-and-password authentication, which is vulnerable to credential theft and identity misuse. This study implements a fingerprint-based biometric authentication system as an identity-verification layer for E-Signature systems, using minutiae-based matching to extract ridge endings and bifurcations from fingerprint images. The system relies on the NIST Biometric Image Software (NBIS), combining the MINDTCT module for minutiae extraction with the Bozorth3 algorithm for matching, and was developed and tested on a Linux Debian environment using a Digital Persona U.are.U 4500 fingerprint sensor. Testing was conducted on 200 genuine-impostor score pairs to determine the optimal matching threshold, followed by a performance evaluation involving 30 respondents across 450 authentication attempts, using Accuracy, Precision, Recall, Specificity, False Acceptance Rate (FAR), False Rejection Rate (FRR), Equal Error Rate (EER), and processing time as evaluation parameters. Results show that a threshold value of 30 provides the best balance between security and usability, yielding a FAR of 0.00% and an FRR of 1.00% at the threshold-determination stage. Full-scale performance testing produced an accuracy of 90.67%, precision of 100.00%, recall of 87.27%, specificity of 100.00%, an operational FAR of 0.00% and FRR of 12.73%, and an average total authentication time of 0.30 seconds. The system successfully rejected all unauthorized fingerprints while operating within real-time processing requirements, demonstrating that NBIS-based fingerprint authentication can strengthen identity verification in Electronic Signature systems.