Scholarlink

<p>Biometric iris image authentication and verification use iris patterns for secure identity confirmation. Current challenges include susceptibility to spoofing, high costs, privacy concerns, and the need for robust systems to handle iris image variations. To address this limitation, a novel approach is proposed that integrates Cyclic Redundancy Checks (CRC) and Error Detection Codes (EDC) directly into the binary iris template during encoding and transmission, enabling detection and correction of bit-level errors caused by noise, compression, or network instability. The system integrates an Adaptive Guided Multi-layer Side Window Box Filter (AGM-SWBF) to reduce overfitting, ASCU-Net for accurate segmentation, and a Gold Rush Optimized Deep Convolutional Sheaf Attention Network (GDCSAN) for feature extraction. Parameter optimization via Gold Rush Optimizer (GRO) enables robust iris template generation, followed by secure AES-chaotic map-based encryption. The proposed method achieves 99.95% accuracy, with FRR of 0.14% and FAR of 0.09%, surpassing recent methods such as HGSSA-BiLSTM, Neuro-Fuzzy, and CNN-SVM by up to 85.71% in FRR and 89% in FAR. Segmentation metrics reach Dice of 0.91 and IoU of 0.83, while template matching accuracy improves to 91.5%, confirming robust extraction and security. Processing time is reduced by 93% over AES, maintaining strong encryption with minimal overhead. Additionally, multimodal evaluation combining iris with fault detection and cybersecurity confirms the system’s scalability, achieving 96.9% accuracy, 97.2%-bit error resilience, and 96.3% fault recovery, outperforming recent multimodal techniques by 20–30%. These results validate the model’s effectiveness in secure, accurate, and fault-tolerant biometric authentication.</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

Cyclic Redundancy Checks and Error Detection Codes with Optimized Attention Based Biometric Iris Image Authentication and Verification

  • 期刊论文

  • 发表日期：

    2026年7月3日

查看全文
  • M. Lavanya,
  • V. Kavitha

摘要

Biometric iris image authentication and verification use iris patterns for secure identity confirmation. Current challenges include susceptibility to spoofing, high costs, privacy concerns, and the need for robust systems to handle iris image variations. To address this limitation, a novel approach is proposed that integrates Cyclic Redundancy Checks (CRC) and Error Detection Codes (EDC) directly into the binary iris template during encoding and transmission, enabling detection and correction of bit-level errors caused by noise, compression, or network instability. The system integrates an Adaptive Guided Multi-layer Side Window Box Filter (AGM-SWBF) to reduce overfitting, ASCU-Net for accurate segmentation, and a Gold Rush Optimized Deep Convolutional Sheaf Attention Network (GDCSAN) for feature extraction. Parameter optimization via Gold Rush Optimizer (GRO) enables robust iris template generation, followed by secure AES-chaotic map-based encryption. The proposed method achieves 99.95% accuracy, with FRR of 0.14% and FAR of 0.09%, surpassing recent methods such as HGSSA-BiLSTM, Neuro-Fuzzy, and CNN-SVM by up to 85.71% in FRR and 89% in FAR. Segmentation metrics reach Dice of 0.91 and IoU of 0.83, while template matching accuracy improves to 91.5%, confirming robust extraction and security. Processing time is reduced by 93% over AES, maintaining strong encryption with minimal overhead. Additionally, multimodal evaluation combining iris with fault detection and cybersecurity confirms the system’s scalability, achieving 96.9% accuracy, 97.2%-bit error resilience, and 96.3% fault recovery, outperforming recent multimodal techniques by 20–30%. These results validate the model’s effectiveness in secure, accurate, and fault-tolerant biometric authentication.