Scholarlink

<p>Fingerprint recognition systems offer a secure, scalable, and non-intrusive method for biometric authentication. However, conventional approaches often struggle with low-quality or partial prints, spoofing attacks, and reduced reliability in real-world conditions. This study presents a biometric cryptosystem that integrates deep learning-based feature extraction with Reed-Solomon (RS) error correction to enhance both accuracy and robustness. The proposed system uses convolutional neural networks (CNNs) to extract hierarchical fingerprint features, which are then discretized and encoded using RS coding for secure and error-tolerant key generation. Fingerprint images from the Sokoto Coventry Fingerprint (SOCOFing) dataset are preprocessed through scaling, normalization, and region of interest (ROI) segmentation before feature extraction. The system is implemented using Python and TensorFlow. Experimental results show perfect authentication performance, achieving a false acceptance rate (FAR) and false rejection rate (FRR) of 0, and a genuine acceptance rate (GAR) of 100%. An area under the ROC curve (AUC) of 1.0 and strong precision-recall metrics further confirm the model’s effectiveness. The cryptosystem also supports efficient enrollment and accurate real-time authentication across multiple users, demonstrating scalability and practical applicability.</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

Enhanced fingerprint authentication: a deep learning and error correction-based biometric cryptosystem

  • 期刊论文

  • 发表日期：

    2025年10月21日

查看全文
  • Netha Merin Mathew,
  • A. Muthukumar

摘要

Fingerprint recognition systems offer a secure, scalable, and non-intrusive method for biometric authentication. However, conventional approaches often struggle with low-quality or partial prints, spoofing attacks, and reduced reliability in real-world conditions. This study presents a biometric cryptosystem that integrates deep learning-based feature extraction with Reed-Solomon (RS) error correction to enhance both accuracy and robustness. The proposed system uses convolutional neural networks (CNNs) to extract hierarchical fingerprint features, which are then discretized and encoded using RS coding for secure and error-tolerant key generation. Fingerprint images from the Sokoto Coventry Fingerprint (SOCOFing) dataset are preprocessed through scaling, normalization, and region of interest (ROI) segmentation before feature extraction. The system is implemented using Python and TensorFlow. Experimental results show perfect authentication performance, achieving a false acceptance rate (FAR) and false rejection rate (FRR) of 0, and a genuine acceptance rate (GAR) of 100%. An area under the ROC curve (AUC) of 1.0 and strong precision-recall metrics further confirm the model’s effectiveness. The cryptosystem also supports efficient enrollment and accurate real-time authentication across multiple users, demonstrating scalability and practical applicability.