Exhaustive Distributed Intrusion Detection System for UAVs Attacks Detection

Abstract

Unmanned Aerial Vehicles (UAVs) are increasingly deployed in modern intelligent systems due to their ability to operate autonomously and collaboratively in complex environments. However, this connectivity introduces significant cybersecurity risks, as communication chan nels and control exchanges become potential targets for malicious attacks. This thesis proposes a distributed intrusion detection framework tailored for UAV net works, where each UAV performs local detection while contributing to a global collaborative decision-making process. Kolmogorov-Arnold Networks (KANs) are adopted as the main learning model due to their ability to efficiently capture nonlinear patterns with low compu tational cost, making them suitable for resource-constrained UAV environments. Three intrusion detection scenarios are studied: network traffic monitoring, MAVLink communication security, and UAVCAN communication security. Experimental results show that the proposed approach achieves strong detection performance while maintaining scala bility and adaptability in dynamic UAV environments.