Machine Learning Models for Intrusion Detection in Unmanned Aerial Vehicles: An Approach to Cybersecurity and Operational Safety

The evolution of Intelligent Transportation Systems (ITS) has significantly advanced aerial transport. The careful integration and optimization of these systems are crucial to ensure safety for both the systems and users. Additionally, cybersecurity is a vital component, especially in urban air mobility, where the coexistence of drones and other aircraft presents new challenges. This study addresses the security of wireless communication in these devices and proposes the development of an Intrusion Detection System (IDS) using machine learning algorithms, emphasizing XGBoost to enhance anomaly identification. The methodology employed includes the development of two detection models: the first is a high-precision model that achieved 99.95 % effectiveness without any false negatives; the second model, while offering broader generalization, maintains an effectiveness of 99.48 % with a 1.4 % occurrence of false negatives. This latter model still outperforms the state-of-the-art model, which achieved 95% accuracy and did not analyze the IDS's false negatives. This work contributes to the detection of intrusions in UAV systems, combining high precision and generalization capabilities. These are essential elements for addressing cybersecurity challenges in the era of urban air mobility.