Scholarlink

<p>The quick growth of the Internet of Things (IoT) connects billions of users and devices in several domains. Like smart homes, healthcare, industrial automation, and transportation. Still, an extensive deployment of IoT devices causes security challenges due to a weak heterogeneous network and limited computation. In an IoT system, cybersecurity is a major concern that affects data secrecy. Still, cyber-attack detection and mitigation in IoT becomes difficult owing to energy constraints, a lack of standard security standards, and limited computational power. To address such issues, this article develops a novel model named Dynamically Weighted Grünwald-Letnikov Derivative-driven Multi-attention Network (DWGL-MANet) for cyber-attack detection and mitigation in IoT. Initially, an IoT system model is developed and then collects input log files. Feature scaling is performed using Variable Stability Scaling, and scaled features are subjected to SelectKBest for reducing feature dimensionality. The detection of cyber-attacks is carried out by DWGL-MANet, in which the learning rules of the Multi-Feature Multi-Attention Network (MFMANet) are developed using Grünwald-Letnikov Derivative and Dynamically Weighted Balanced loss. Finally, cyber-attack mitigation in IoT is carried out. Furthermore, developed DWGL-MANet yields the optimal accuracy, True Positive Rate (TPR), and True Negative Rate (TNR) of 96.36%, 96.42%, and 96.51%.</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

Dynamically weighted Grünwald-Letnikov derivative-driven multi-attention network for cyber-attack detection and mitigation in Internet of Things

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月19日

查看全文
  • Meghana G. Raj

摘要

The quick growth of the Internet of Things (IoT) connects billions of users and devices in several domains. Like smart homes, healthcare, industrial automation, and transportation. Still, an extensive deployment of IoT devices causes security challenges due to a weak heterogeneous network and limited computation. In an IoT system, cybersecurity is a major concern that affects data secrecy. Still, cyber-attack detection and mitigation in IoT becomes difficult owing to energy constraints, a lack of standard security standards, and limited computational power. To address such issues, this article develops a novel model named Dynamically Weighted Grünwald-Letnikov Derivative-driven Multi-attention Network (DWGL-MANet) for cyber-attack detection and mitigation in IoT. Initially, an IoT system model is developed and then collects input log files. Feature scaling is performed using Variable Stability Scaling, and scaled features are subjected to SelectKBest for reducing feature dimensionality. The detection of cyber-attacks is carried out by DWGL-MANet, in which the learning rules of the Multi-Feature Multi-Attention Network (MFMANet) are developed using Grünwald-Letnikov Derivative and Dynamically Weighted Balanced loss. Finally, cyber-attack mitigation in IoT is carried out. Furthermore, developed DWGL-MANet yields the optimal accuracy, True Positive Rate (TPR), and True Negative Rate (TNR) of 96.36%, 96.42%, and 96.51%.