Scholarlink

<p>The widespread integration of smart devices across sectors such as healthcare, manufacturing, and intelligent transportation has positioned the Internet of Things (IoT) as a cornerstone of modern digital infrastructure. However, its pervasive interconnectivity increases exposure to cyber threats, making timely and accurate anomaly detection essential for ensuring network resilience. This study introduces IoTLSDT, a hybrid anomaly detection framework that integrates the temporal sequence learning capabilities of Long Short-Term Memory (LSTM) networks with the interpretability of Decision Trees (DT). The model leverages LSTM Softmax probability outputs as transparent inputs to the DT, effectively combining deep temporal representation with rule-based reasoning. IoTLSDT was evaluated under time-aware validation on three heterogeneous IoT datasets, IoT-23, DAD, and CICIoV-2024, covering diverse network protocols and attack types. Experimental results show that IoTLSDT achieves F1 Score of 0.87, 0.99, and 0.81, respectively, outperforming recent state-of-the-art models such as SELIDS (0.83), Hybrid XGBoost (0.91), and Blockchain-enabled IDS (0.86). These results confirm that IoTLSDT delivers superior detection accuracy, enhanced temporal robustness, and full model explainability while maintaining lightweight inference. The framework thus offers a scalable and interpretable solution for real-time anomaly detection in complex, resource-constrained IoT environments.</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

IoTLSDT hybrid method for anomaly detection in internet of things environments

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月2日

查看全文
  • Dan Ali,
  • Dagogo Orifama,
  • Olatunde Olaleye,
  • Benedict Onochie Ibe,
  • Ayodeji Akeem Ajani,
  • Oluseun Damilola Oyeleke

摘要

The widespread integration of smart devices across sectors such as healthcare, manufacturing, and intelligent transportation has positioned the Internet of Things (IoT) as a cornerstone of modern digital infrastructure. However, its pervasive interconnectivity increases exposure to cyber threats, making timely and accurate anomaly detection essential for ensuring network resilience. This study introduces IoTLSDT, a hybrid anomaly detection framework that integrates the temporal sequence learning capabilities of Long Short-Term Memory (LSTM) networks with the interpretability of Decision Trees (DT). The model leverages LSTM Softmax probability outputs as transparent inputs to the DT, effectively combining deep temporal representation with rule-based reasoning. IoTLSDT was evaluated under time-aware validation on three heterogeneous IoT datasets, IoT-23, DAD, and CICIoV-2024, covering diverse network protocols and attack types. Experimental results show that IoTLSDT achieves F1 Score of 0.87, 0.99, and 0.81, respectively, outperforming recent state-of-the-art models such as SELIDS (0.83), Hybrid XGBoost (0.91), and Blockchain-enabled IDS (0.86). These results confirm that IoTLSDT delivers superior detection accuracy, enhanced temporal robustness, and full model explainability while maintaining lightweight inference. The framework thus offers a scalable and interpretable solution for real-time anomaly detection in complex, resource-constrained IoT environments.