Evaluation of Machine Learning Models for Attack Detection in Unmanned Aerial Vehicle Networks

Nowadays, unmanned aerial vehicles (UAVs) are increasingly utilized in various civil and military applications, highlighting the growing need for robust security in UAV networks. Cyberattacks on these networks can lead to operational disruptions and the loss of critical information. This study evaluates five machine learning models—Random Forest (RF), CatBoost, XGBoost, AdaBoost, and Artificial Neural Networks (ANN)—for detecting attacks on UAV networks using the CICIOT2023 (Canadian Institute for Cybersecurity Internet of Things 2023) dataset. Performance metrics such as accuracy, precision, sensitivity, and F1 score were used to assess these models. Among them, CatBoost demonstrated superior performance, achieving the highest accuracy and the fastest prediction time of 6.487 seconds, making it particularly advantageous for real-time attack detection. This study underscores the effectiveness of CatBoost in both accuracy and efficiency, positioning it as an ideal choice for enhancing UAV network security. The findings contribute to addressing cybersecurity vulnerabilities in UAV networks and support the development of more secure network infrastructures.