XGBoost and Deep Learning Based-Federated Learning for DDoS Attack Detection in 5G Core Network VNFs

Abstract

Distributed Denial-of-Service (DDoS) attacks can employ cloud network zombies to compromise the availability of 5G network services and hinder the ability of telecommunication service providers (TSPs) to deliver promised service level agreements (SLAs), which can lead to potential losses for both TSPs and their clients. In this study, a technique for mitigating DDoS attacks in 5G core network Virtual Network Functions (VNFs) was proposed. VNFs are cloud-based network functions that provide 5G services. First, eXtreme Gradient Boosting (XGBoost) is used to extract relevant features, and then the proposed hybrid deep neural network uses the XGBoost-extracted features for DDoS attack detection and mitigation. To address data privacy in heterogeneous 5G networks, where VNFs can be provided and hosted in different cloud computing systems, Federated Learning (FL) is used for the proposed model training. The anti-DDoS framework was evaluated through a simulation using the CICDDoS2019 dataset with 10 VNFs for FL. Through extensive experimentation and evaluation, the results demonstrated promising outcomes characterized by high accuracy, low false positive rate, and minimal detection time. The proposed solution offers a robust defense mechanism against evolving DDoS threats that can target 5G core networks, thereby ensuring the availability of critical network infrastructure.