Abstract
Nowadays, biometric confirmation systems are utilized for security applications such as verification and identification. There are various biometric modalities such as fingerprints, face recognition, and iris scans. Biometric systems are superior to PIN and password-based systems because the latter can be easily stolen or forgotten, whereas biometric traits are unique and difficult to replicate or forget. Among biometric modalities, fingerprint recognition is widely employed for security purposes due to the distinctiveness of each individual's fingerprint. However, fingerprint biometric systems encounter challenges, such as the reproduction of fake fingerprints using materials like silicon, which can potentially bypass security measures. This issue is considered a significant problem in fingerprint systems.This paper proposes a new software-based method called Dynamic Local Ternary Pattern (DLTP) for fingerprint liveness detection, employing a machine learning classifier, specifically Support Vector Machine (SVM), to distinguish between live and fake fingerprints. Various experiments were conducted using DLTP and state-of-the-art texture descriptors. The results obtained from DLTP demonstrated optimal accuracy, clearly surpassing those achieved by the Local Binary Pattern (LBP) and Local Ternary Pattern (LTP) texture descriptors reported in previous studies.