Machine Learning-based Intrusion Detection and Explainable AI in IoT Networks

Authors

Amber Hoenig, Computational Data Science and Engineering, College of Engineering, North Carolina Agricultural and Technical State University
Laila Holland, Computational Data Science and Engineering, College of Engineering, North Carolina Agricultural and Technical State University
Alexander Nanor

Files

Description

Machine learning-based intrusion detection systems (IDS) are one of the most vital and widely used defenses for Internet of Things networks. However, the ever-changing landscape of cyberattacks and the lack of explainability of most ML and deep learning methods leaves these systems vulnerable to intrusions, inaccuracies, compromised data, and new attacks. This research implements machine learning, deep learning, and explainable AI to assess the strengths and drawbacks of each method. Our results show that the inherently explainable model of decision trees (DT) provides excellent results similar to the highest performing models of LSTM and GRU. Achieving results within 2% of the best performing models and offering by far the lowest training and testing times of all models, Decision Trees demonstrates its viability as a computationally efficient, high-performing, inherently explainable model for IoT intrusion detection systems

Publication Date

4-1-2025

Keywords

Explainable Artificial Intelligence, Internet of Things, Machine Learning, Deep Learning

Recommended Citation

Hoenig, Amber; Holland, Laila; and Nanor, Alexander, "Machine Learning-based Intrusion Detection and Explainable AI in IoT Networks" (2025). 2025 Graduate Student Research Symposium. 161.
https://digital.library.ncat.edu/gradresearchsymposium25/161