<![CDATA[The increasing integration of smart devices into daily life has made the Internet of Things (IoT) essential across sectors such as manufacturing, transportation, healthcare, and smart homes. While IoT offers substantial benefits in automation and real-time monitoring, its pervasive connectivity exposes networks to significant security threats. Timely detection of anomalies is therefore critical to ensuring system resilience. This study presents IoTLSDT, a novel hybrid anomaly detection model that combines the temporal learning strengths of Long Short-Term Memory (LSTM) networks with the interpretability of Decision Trees. The model was trained and evaluated on three diverse and publicly available IoT datasets, including CICIoT2024, DAD, and IoT-23, which cover various attack types and traffic behaviours. Unlike existing methods, IoTLSDT utilises SoftMax probability outputs from the LSTM as input features for the Decision Tree, enhancing both performance and explainability. Experimental results demonstrate that IoTLSDT consistently outperforms conventional machine learning models, achieving classification accuracies ranging from 86% to 99% across all datasets. These results suggest that the proposed model is a robust and scalable solution for real-time anomaly detection in heterogeneous IoT environments.]]>
Copyrights © 2025
