Scholarlink

<p>Vein biometric authentication systems are crucial in military and forensic applications, where security remains a primary concern. The potential compromise of stored biometric data within databases poses a significant security risk. To address this problem, a hybrid biometric template protection method utilizing chaotic encryption and innovative perturbation algorithms has been introduced. The vein images are subjected to preprocessing, and vein line patterns are extracted utilizing line-tracking methods in conjunction with curvature analysis. The extracted features are integrated employing a feature fusion technique and preserved through an innovative perturbation-based encryption method. The encrypted images are subjected to thorough security evaluations, encompassing correlation analysis, pixel change rate (NPCR) testing, histogram examination, and entropy assessment. The proposed multimodal system exhibits an accuracy of 99.85% and an Equal Error Rate of 0.017 when utilizing the ResNet deep learning mode.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A robust biometric system using wrist and dorsal vein images for person authentication

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月8日

查看全文
  • V. Gurunathan,
  • R. Sudhakar

摘要

Vein biometric authentication systems are crucial in military and forensic applications, where security remains a primary concern. The potential compromise of stored biometric data within databases poses a significant security risk. To address this problem, a hybrid biometric template protection method utilizing chaotic encryption and innovative perturbation algorithms has been introduced. The vein images are subjected to preprocessing, and vein line patterns are extracted utilizing line-tracking methods in conjunction with curvature analysis. The extracted features are integrated employing a feature fusion technique and preserved through an innovative perturbation-based encryption method. The encrypted images are subjected to thorough security evaluations, encompassing correlation analysis, pixel change rate (NPCR) testing, histogram examination, and entropy assessment. The proposed multimodal system exhibits an accuracy of 99.85% and an Equal Error Rate of 0.017 when utilizing the ResNet deep learning mode.