| A Systematic Modelling of Intrusion Detection in IoT Using a Hybrid Machine Learning Approach | |
|
MSRDG International Journal of Computer Scientific Technology & Electronics Engineering
© 2026 by MSRDG IJCSTEE Journal Volume 2 Issue 3
Year of Publication: 2026 |
Paper Download Article ID MSRDG-IJCSTEE-V2I3P102 |
|
Abstract: The rapid proliferation of Internet of Things (IoT) devices has expanded the attack surface of modern networks, making reliable intrusion detection systems (IDS) essential for securing resource-constrained and heterogeneous IoT environments. Traditional signature-based and single-model machine learning IDS approaches often struggle to generalize across diverse attack types and suffer from high false-positive rates when applied to imbalanced, high-dimensional IoT traffic data. This paper presents a systematic modelling of a Hybrid Machine Learning Intrusion Detection framework for IoT (H-IDS-IoT) that combines a deep-learning branch (convolutional neural network for spatial feature extraction followed by a long short-term memory network for temporal sequence modelling) with a classical ensemble branch (Random Forest and XGBoost) fused through a stacking meta-learner. The framework further incorporates class-balancing (SMOTE) and mutual-information/PCA-based feature selection to address the class imbalance and dimensionality challenges characteristic of IoT network traffic. We describe the proposed architecture, the end-to-end data processing and training workflow, and an evaluation methodology based on standard classification metrics (accuracy, precision, recall, F1-score, false-positive rate, and AUC) benchmarked against individual classifiers (SVM, Random Forest, CNN, LSTM). A representative evaluation design and illustrative result visualizations, including comparative performance charts, ROC curves, and a confusion matrix, are presented to demonstrate how the framework's detection capability would be assessed on a real IoT intrusion dataset. The proposed hybrid approach is intended to improve detection accuracy and reduce false alarms relative to single-model baselines in multi-class IoT attack classification. |
|
| Keywords: Internet of Things, Intrusion Detection System, Hybrid Machine Learning, Deep Learning, Ensemble Learning, Network Security, Feature Selection, Anomaly Detection | |
