Scholarlink

<p>Nowadays, the Internet of Things (IoT) is proliferating, bringing worldwide alarm for the security of interconnected networks, which requires an effective threat detection model to lessen cyber threats. Traditional threat detection models often struggle to identify complex attacks happening in an IoT network due to heterogeneous traffic features, data imbalance, and limited feature learning. This study aims to design an accurate threat detection in an IoT network using an attention-based deep learning (DL) algorithm that enhances spatial and temporal feature learning. Also, data-balancing and feature selection methods are employed in the research to improve accuracy and robustness. The proposed model was tested on two widely used datasets: IoT-23 and N-BaIoT, which provide realistic, diverse data on IoT network traffic and botnet attacks that are crucial for developing robust security systems. The experimental findings demonstrated that our model shows remarkable performance for threat detection compared to existing techniques by reaching a maximum accuracy of 99.93% and 99.86% for the corresponding IoT-23 and N-BaIoT datasets. These results highlight the efficiency of combining attention-included DL-based feature learning with an optimized temporal learning mechanism to improve generalization and reduce misclassification in thread detection.</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

Enhancing IoT Network Security Through an Advanced Deep Learning-Powered Threat Detection System

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • V. Vaissnave,
  • Guruprakash Baskaran,
  • Suja Cherukullapurath Mana,
  • K. Abirami

摘要

Nowadays, the Internet of Things (IoT) is proliferating, bringing worldwide alarm for the security of interconnected networks, which requires an effective threat detection model to lessen cyber threats. Traditional threat detection models often struggle to identify complex attacks happening in an IoT network due to heterogeneous traffic features, data imbalance, and limited feature learning. This study aims to design an accurate threat detection in an IoT network using an attention-based deep learning (DL) algorithm that enhances spatial and temporal feature learning. Also, data-balancing and feature selection methods are employed in the research to improve accuracy and robustness. The proposed model was tested on two widely used datasets: IoT-23 and N-BaIoT, which provide realistic, diverse data on IoT network traffic and botnet attacks that are crucial for developing robust security systems. The experimental findings demonstrated that our model shows remarkable performance for threat detection compared to existing techniques by reaching a maximum accuracy of 99.93% and 99.86% for the corresponding IoT-23 and N-BaIoT datasets. These results highlight the efficiency of combining attention-included DL-based feature learning with an optimized temporal learning mechanism to improve generalization and reduce misclassification in thread detection.