Abstract
The goal of this paper is to improve the detection efficiency of Denial-of-Service (DoS) attacks in cellular networks. The study uses the TON_IoT dataset, which has a lot of different kinds of data, such as network traffic from a realistic medium-scale network, operating system logs, and telemetry from Internet of Things (IoT) and Industrial Internet of Things (IIoT) services. We conducted a comprehensive analysis of various feature selection methods and machine learning models to achieve the research objectives. We applied the statistical feature selection method to identify the most informative characteristics of network traffic associated with the selected attack type. We employed multiple machine learning algorithms, including random forest, logistic regression, and support vector machines (SVM), for traffic classification. We analyzed the impact of training dataset size and the significance of individual features on classification accuracy. The experimental evaluation identified the optimal combination of feature selection method and machine learning model, which provided the highest accuracy in detecting both DoS and DDoS attacks. The obtained results demonstrate the high effectiveness of the proposed approach and open up new possibilities for developing more robust intrusion detection systems and protecting cellular networks.