Advanced Threat Detection in IoT Networks using Hyperparameter-Tuned Machine Learning Models

Mirle Nataraja, Tejasvi (2024) Advanced Threat Detection in IoT Networks using Hyperparameter-Tuned Machine Learning Models. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (648kB) | Preview
Preview	PDF (Configuration Manual) Download (368kB) | Preview

Abstract

The use of IoT (Internet of Things) devices has brought several new forms of security risks which cause growth of malware that targets IoT networks. Current techniques for intrusion detection involve a IoT of interaction and may thus not be very efficient in handling new kind of threats.

The purpose of the research is to propose the IoT-2023 based efficient machine learning models for improving the identification of the illicit actions in the IoT networks. The proposed method combines feature selection and classification techniques which are useful for identifying the features that best characterize the malicious and benign network traffics. To get the best from the machine learning models developed hyperparameter tuning is utilized.

Machine learning classifiers have been developed such as Logistic regression, Decision tree, Decision stump, Random Forest, Naïve Bayes, K Nearest neighbours (KNN), Support Vector Machines (SVM), Multi Layer Perception (MLP), Gradient Boosting (GB) and Extreme Gradient Boosting (XGB). All these models work for have been made to work with binary and multi class classification tasks. In addition hyperparameter tuning is done for all the models. There is a significant improvement in the accuracy of the model.

Boosting algorithms like Gradient Boosting (GB) and Extreme gradient boosting (XGB) are working greatly with binary classification of the attack types with 95% accuracy and 95% F1 score each. Decision tree and random forest algorithm is capable to handle multi class classification in a better way with 84% accuracy each. Although hyperparameter tuning is responsible for model improvement, it has no compatibility with models like Naïve Bayes and Decision Stump models. With this accuracy of the model receded with complex classification tasks like 33% accuracy for decision stump and 67% accuracy for Naïve Bayes. With these studies we were able to reach the research objectives.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Milosavljevic, Vladimir UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Telecommunications > Computer networks > Internet of things Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	03 Sep 2025 14:09
Last Modified:	03 Sep 2025 14:09
URI:	https://norma.ncirl.ie/id/eprint/8749

Actions (login required)

View Item