Unmanned aerial vehicles are increasingly becoming part of interconnected cyber-physical systems, with applications in smart traffic management, infrastructure inspection, logistics, and emergency response. Yet their safe and trustworthy operation remains heavily dependent on positioning, timing, and communication mechanisms that are vulnerable to attack or failure. This challenge is particularly acute in GPS-denied or degraded environments, such as indoor spaces, dense urban areas, or hostile settings affected by spoofing and jamming. In this talk, I will present part of my research agenda on UAV cybersecurity through two recent studies on UAV security. The first examines protocol-level weaknesses in MAVLink 2.0, showing how GPS time manipulation can undermine message freshness logic, create replay conditions, and trigger denial-of-service behaviour even without compromise of the signing key. The second presents a deep learning-based framework for detecting GPS spoofing attacks using onboard GPS-derived data, combining BiLSTM, attention, and convolutional layers to identify anomalous flight behaviour in realistic attack scenarios. Together, these studies show that UAV security must be addressed across multiple layers, from navigation signals and time synchronisation to protocol design and anomaly detection. I will conclude by discussing how these findings inform the design of resilient UAVs for GPS-denied environments and what they imply for the security of future interconnected cyber-physical systems.
Dr. Alma Oracevic is a Senior Lecturer in Cybersecurity at the University of Bristol and a member of the Bristol Cyber Security Group. Her research focuses on UAV security, cyber-physical system resilience, and trustworthy autonomous systems, with particular emphasis on GPS-denied navigation, vulnerability analysis, and secure communication for drones. She currently holds an RAEng / Leverhulme Trust Research Fellowship for her project DroneShield: Cyber Vulnerability Analysis in GPS-Denied Environments.