Scholarlink

<p>Biometrics has captured the interest of researchers in recent years. Biometric systems depend on feature extraction and storage in databases. This makes the systems vulnerable to attacks. To save these systems from attacks, it is possible to build cancelable biometric systems through processing of the optical biometric templates. The cancelable biometric systems depend on unreal templates that can be obtained from either a non-invertible transform or an encryption algorithm on the signal or feature level. In this paper, we use two proposed realizations of a cascaded structure to generate cancelable biometrics. The first one is based on the adaptive filter and the comb filter in consequence, while the second one is based on the comb filter and then the adaptive filter. The cascaded structure is used to generate the cancelable patterns for faces and fingerprints, and then we compare the realizations with each other. The combined filters perform the deformation process. We observe that for different comb filter orders (<i>L</i>), there is low correlation between the original and generated biometrics. The different filter orders tested in this study include 6, 8, 10, and 12. The proposed system is assessed using Equal Error Rate (EER), False Accept Rate (FAR), False Reject Rate (FRR), and Area under Receiver Operating Characteristic (ROC) curve. The first proposal (adaptive and then comb filter) is preferred for faces and fingerprints. There is no intersection between genuine and imposter distribution curves, and AROC = 1. If the filter order increases, the results get better. At order <i>L</i> = 12 in case face images, EER = 1.2019&#xa0;×&#xa0;10<sup>-07</sup>, FAR = 2.9074&#xa0;×&#xa0;10<sup>-09</sup> and FRR = 3.7155&#xa0;×&#xa0;10<sup>-07</sup>.</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

Utilization of optical techniques for privacy preservation

  • 期刊论文

  • 发表日期：

    2025年4月23日

查看全文
  • Mohamed AbdAlRaheem,
  • Fathi E. Abd El-Samie,
  • Walid El-Shafai,
  • Osama A. Omer

摘要

Biometrics has captured the interest of researchers in recent years. Biometric systems depend on feature extraction and storage in databases. This makes the systems vulnerable to attacks. To save these systems from attacks, it is possible to build cancelable biometric systems through processing of the optical biometric templates. The cancelable biometric systems depend on unreal templates that can be obtained from either a non-invertible transform or an encryption algorithm on the signal or feature level. In this paper, we use two proposed realizations of a cascaded structure to generate cancelable biometrics. The first one is based on the adaptive filter and the comb filter in consequence, while the second one is based on the comb filter and then the adaptive filter. The cascaded structure is used to generate the cancelable patterns for faces and fingerprints, and then we compare the realizations with each other. The combined filters perform the deformation process. We observe that for different comb filter orders (L), there is low correlation between the original and generated biometrics. The different filter orders tested in this study include 6, 8, 10, and 12. The proposed system is assessed using Equal Error Rate (EER), False Accept Rate (FAR), False Reject Rate (FRR), and Area under Receiver Operating Characteristic (ROC) curve. The first proposal (adaptive and then comb filter) is preferred for faces and fingerprints. There is no intersection between genuine and imposter distribution curves, and AROC = 1. If the filter order increases, the results get better. At order L = 12 in case face images, EER = 1.2019 × 10-07, FAR = 2.9074 × 10-09 and FRR = 3.7155 × 10-07.