Scholarlink

Internet of Things (IoT) gadgets connect and simplify life in several ways. The independence of these technologies poses some issues, such as the protection of security and privacy from malicious and compromised nodes inside the network. To overcome this complication, Enhancing IoT Security Through Hierarchical Message-Passing Graph Neural Networks: A Trust-Driven Strategy for Identifying Malicious Nodes (HMPGNN-Trust-IoT) is proposed. These issues are addressed by the proposed HMPGNN-Trust-IoT solution, which develops a mechanism for recognizing malicious and compromised nodes utilizing trust-based deep learning. With the help of the dataset provided, HMPGNN-Trust develops a worldwide model to predict the IoT node’s irrational behavior. A global dataset comprising 19 trust parameters on three main categories including knowledge, experience, and reputation are used in the proposed HMPGNN-Trust-IoT technique. The data are fed to pre-processing. In preprocessing, it cleans and normalizes the data in the dataset by utilizing a Nanoplasmonic Ultra Wideband Band Pass Filter (NUWBF). HMPGNN-Trust uses the idea of communities using dedicated servers to divide the dataset into smaller portions for effective training to lessen the computational strain. Finally, it classifies the nodes by utilizing a Hierarchical Message-Passing Graph Neural Network (HMPGNN) into benign and malicious nodes. The proposed HMPGNN-Trust-IoT is implemented using Python. To identify Trust Management in IOT performance metrics like precision, accuracy, and F1-score are considered. Performance of the HMPGNN-Trust-IoT approach attains 24.11%, 28.56%, and 22.73% high accuracy, 21.89%, 23.04%, and 9.51% high precision, 25.289%, 15.35%, and 19.91% higher F1-Score compared with existing methods such as the Securing IoT along Deep Federated Learning: A Trust-based Malicious Node Identification method (TM-ANN -IoT), the Trust-driven reinforcement selection technique in federated learning on Internet of Things devices (TM-SDQN-IoT) and the role of machine learning methods in internet of things-based cloud applications (TM-SVM-IoT) models respectively.

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 Security Through Hierarchical Message-Passing Graph Neural Networks: A Trust-Driven Strategy for Identifying Malicious Nodes

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • C. Senthil Kumar,
  • R. Vijay Anand

摘要

Internet of Things (IoT) gadgets connect and simplify life in several ways. The independence of these technologies poses some issues, such as the protection of security and privacy from malicious and compromised nodes inside the network. To overcome this complication, Enhancing IoT Security Through Hierarchical Message-Passing Graph Neural Networks: A Trust-Driven Strategy for Identifying Malicious Nodes (HMPGNN-Trust-IoT) is proposed. These issues are addressed by the proposed HMPGNN-Trust-IoT solution, which develops a mechanism for recognizing malicious and compromised nodes utilizing trust-based deep learning. With the help of the dataset provided, HMPGNN-Trust develops a worldwide model to predict the IoT node’s irrational behavior. A global dataset comprising 19 trust parameters on three main categories including knowledge, experience, and reputation are used in the proposed HMPGNN-Trust-IoT technique. The data are fed to pre-processing. In preprocessing, it cleans and normalizes the data in the dataset by utilizing a Nanoplasmonic Ultra Wideband Band Pass Filter (NUWBF). HMPGNN-Trust uses the idea of communities using dedicated servers to divide the dataset into smaller portions for effective training to lessen the computational strain. Finally, it classifies the nodes by utilizing a Hierarchical Message-Passing Graph Neural Network (HMPGNN) into benign and malicious nodes. The proposed HMPGNN-Trust-IoT is implemented using Python. To identify Trust Management in IOT performance metrics like precision, accuracy, and F1-score are considered. Performance of the HMPGNN-Trust-IoT approach attains 24.11%, 28.56%, and 22.73% high accuracy, 21.89%, 23.04%, and 9.51% high precision, 25.289%, 15.35%, and 19.91% higher F1-Score compared with existing methods such as the Securing IoT along Deep Federated Learning: A Trust-based Malicious Node Identification method (TM-ANN -IoT), the Trust-driven reinforcement selection technique in federated learning on Internet of Things devices (TM-SDQN-IoT) and the role of machine learning methods in internet of things-based cloud applications (TM-SVM-IoT) models respectively.