Detection of DoS Attacks in IoT Networks: Comparative Study of Selected Supervised Machine Learning Algorithms

The article presents the application of selected supervised machine learning algorithms for detecting Denial of Service (DoS) attacks in Internet of Things (IoT) networks. Data from the TON_IoT dataset, simulating an extensive IoT environment, was utilized and subjected to a multi – stage process of preprocessing and extraction of significant features. The prepared test environment enabled a comprehensive comparison of 5 machine learning algorithms: Decision Tree, Random Forest, XGBoost, k – Nearest Neighbors, and Multilayer Perceptron. Popular confusion matrix – based metrics were employed to evaluate the effectiveness of DoS attack detection, and model training/prediction times were measured. The XGBoost algorithm achieved the highest classification accuracy while simultaneously exhibiting the lowest processing time.