DoS Attack Detection in 5G Core Network using Machine Learning

Abdel Fadyl Chabi, João Vitor da S. Campos, Vivianne de Aquino Rodrigues, Matheus Fontinele de Aguiar

DOI: 10.14209/sbrt.2025.1571157302

Evento: XLIII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2025)

Keywords: 5G networks Security DoS attack Machine learning

Abstract

Contrary to previous mobile networks, 5G architecture is service based and utilize Network Functions (NFs) to manage connectivity, security, data management, and other functionalities of the network. This kind of architecture allow 5G networks to be more scalable, flexible and eficient than the legacy networks, however, service based networks also introduces a series of vulnerabilities. As 5G networks evolve, security becomes as much of a concern as network performance. 5G networks allow the introduction of machine learning techniques, which can improve network performance, as well as ensure security through the implementation of machine learning models on both base station and User Equipment (UE) side. In this paper we use five machine learning algorithm, namely Random Forest, XGboost, Decision Tree, Multi-Layer Perceptron (MLP) and LightGBM to detect Denial of Service (DoS) attack in 5G core network. Friedmann and Nemenyi tests have been performed to compare the models performance and elect the best model among the five implemented. Random Forest, XGBoost and Decision Tree are elected the best model for Dos attack detection problem, when comparing the metrics and statistical results obtained.

Download