Deep Graph Learning for DDoS Detection and Multi-Class Classification IDS

Critical infrastructure systems have been preyed on by cyber criminals that target to disrupt their operations and national security. Among the most nefarious attacks, the Distributed Denial of Service (DDoS) attack is wreaking havoc on the Telecommunications sector. This paper invests in the vision that Artificial Intelligence (AI) plays an important role in shoring up the cybersecurity of critical infrastructure providers by detecting and classifying malicious engagements. In this respect, we propose an efficient and dependable DDoS specialized intrusion section system (IDS). The proposed system is empowered by Graph Convolutional Networks (GCN), a deep learning technique, which is capable of capturing the topological and statistical information between the attack network and the victim network. The results show that the proposed GCN IDS can detect and classify multiple variations of DoS with a high confidence level.

I. Introduction

Critical infrastructure providers across different sectors such as communications, healthcare and public health sector, finan-cial, and energy, to name a few, have been victimized by cybercriminal organizations systematically. Among the most nefarious cyber attacks, is the Distributed Denial of Service (DDoS) that wreaks havoc on the availability of the targeted systems and networks precluding legitimate access from taking place. Recent statistics on DDoS attacks have highlighted a massive increase of 387% in the number of attacks when contrasting the second quarter (Q2) and the first one (Q1) of 2023 where the telecommunication sector alone accounts for approximately 50% of these malicious campaigns [1].