Enhancing UAV Network Security: Machine Learning for DDoS Attack Detection

Assaults on network components and devices are a major threat to network services. Therefore, new technical solutions need to be developed to counter these attacks and improve the efficiency of attack detection systems. This paper presents a Distributed Denial of Service (DDoS) attack detection algorithm for mobile Unmanned Aerial Vehicles (UAV) networks. We consider a Software Defined Networks architecture and propose a Machine Learning (ML) based attack detection module. By incorporating novel features such as flow count, speed of source IP, source IP address entropy, destination IP address entropy, etc., in our ML-based model, we show that the detection performance significantly increases in the UAV networks. ML algorithms used are Random Forest, Multi-Layer Perceptron, and Support Vector Machine. The performance has been evaluated under varying traffic types (TCP, UDP, and ICMP) and attack traffic intensities (light and heavy). In comparison to the state-of-the-art schemes having an accuracy of approximately 95%, our proposed scheme achieves an average accuracy of around 99%. Our current work paves the way for the development of an efficient attack mitigation system in mobile UAV networks.