EFFICIENT IoT INTRUSION DETECTION USING MUTUAL INFORMATION AND ENSEMBLE MACHINE LEARNING
Journal of Systematic and Modern Science Research · 02.07.2026
DOI: 10.70382/bejsmsr.v12i9.025 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Systematic and Modern Science Research, online erschienen am 02.07.2026. Verfasst von ASEGUNLOLUWA E. BABALOLA und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The increasing deployment of Internet of Things devices has expanded the attack surface of connected environments, creating a need for intrusion detection systems that are both accurate and computationally efficient. This study investigates the use of Mutual Information-based feature selection and machine learning for multiclass IoT intrusion detection using the CICIoT2023 dataset. The original attack labels are reorganized into eight broader traffic categories, while Mutual Information ranks the network traffic variables and identifies compact feature subsets. Logistic Regression, Multilayer Perceptron, Extra Trees and Random Forest are evaluated using the complete feature set and selected subsets of the highest-ranked variables. The results show that the ensemble tree-based models outperform the linear and neural-network approaches. Random Forest trained with the top 10 features provides the best balance between overall accuracy, class-balanced performance and computational efficiency. The reduced configuration also improves the recognition of rare attack categories while maintaining strong detection of dominant attacks. Also, it shortens model training and prediction times compared with the full-feature configuration. The findings demonstrate that Mutual Information and Random Forest can support the development of lightweight intrusion detection systems for resource-constrained IoT environments. The approach is applicable to smart homes, industrial IoT networks, healthcare devices, smart cities and other connected systems requiring real-time monitoring and early identification of malicious traffic.
Zitieren
APA
BABALOLA, A. E., UDEOGU, N. M., & ADEOLA, P. (2026). EFFICIENT IoT INTRUSION DETECTION USING MUTUAL INFORMATION AND ENSEMBLE MACHINE LEARNING. Journal of Systematic and Modern Science Research. https://doi.org/10.70382/bejsmsr.v12i9.025
Vancouver
BABALOLA AE, UDEOGU NM, ADEOLA P. EFFICIENT IoT INTRUSION DETECTION USING MUTUAL INFORMATION AND ENSEMBLE MACHINE LEARNING. Journal of Systematic and Modern Science Research. 2026. doi:10.70382/bejsmsr.v12i9.025
BibTeX
@article{BABALOLA2026,
author = {BABALOLA, ASEGUNLOLUWA E. and UDEOGU, NDUDIM M. and ADEOLA, PREYE},
title = {EFFICIENT IoT INTRUSION DETECTION USING MUTUAL INFORMATION AND ENSEMBLE MACHINE LEARNING},
journal = {Journal of Systematic and Modern Science Research},
year = {2026},
doi = {10.70382/bejsmsr.v12i9.025}
}
EUVIMED