Scholarlink

<p>The usage of the internet and other handheld smart devices promotes the theft of users’ biometric traits, which frequently results in a loss of privacy. Protecting biometric template data is essential to prevent loss of user privacy and identification. An effective and efficient approach is required that focuses on the security and privacy concerns associated with biometric traits. To cope with this issue, we propose a convolutional autoencoder and a winner-take-all hashing-based cancelable biometric template. For each image, the latent-space vector representation is computed, and the features are extracted using the convolutional autoencoder. The generated biometric feature vector is further enhanced with random noise to increase the discriminative power and enable the revocability of biometric templates. When computing similarity, winner-take-all hashing is used to determine the relevant features and effectively ignores the rest of the information. The performance of the suggested approach is evaluated using the fingerprint data sets FVC2002-DB1 and DB2. The results of the experiments and the security analysis support the effectiveness of the suggested approach.</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

Privacy-Preserving fingerprint authentication based on cancelable biometrics

  • 期刊论文

  • 发表日期：

    2026年4月22日

查看全文
  • Ayesha S. Shaikh,
  • Vibha D. Patel

摘要

The usage of the internet and other handheld smart devices promotes the theft of users’ biometric traits, which frequently results in a loss of privacy. Protecting biometric template data is essential to prevent loss of user privacy and identification. An effective and efficient approach is required that focuses on the security and privacy concerns associated with biometric traits. To cope with this issue, we propose a convolutional autoencoder and a winner-take-all hashing-based cancelable biometric template. For each image, the latent-space vector representation is computed, and the features are extracted using the convolutional autoencoder. The generated biometric feature vector is further enhanced with random noise to increase the discriminative power and enable the revocability of biometric templates. When computing similarity, winner-take-all hashing is used to determine the relevant features and effectively ignores the rest of the information. The performance of the suggested approach is evaluated using the fingerprint data sets FVC2002-DB1 and DB2. The results of the experiments and the security analysis support the effectiveness of the suggested approach.