Synthetic threat Dataset Generation by UAV fleet Simulation

In the coming years, institutions, such as law enforcement, rescue agencies, and companies, will increasingly use drone fleets to support inspection, logistics, and surveillance applications, in scenarios that may also be highly sensitive for managed data, performed tasks, or the contexts in which they operate. Security-by-design and machine learning-based models may represent the main paradigms for the design of systems resistant to cyber attacks. However, the lack of open datasets and simulation frameworks may represent the main challenge in this direction. Therefore, this paper aims to propose an open source software stack and a schema for creating synthetic standard missions, attacks, and component failures, for the generation of scenarios and data useful for addressing security threats against drone fleets. The produced data set may be used to implement machine learning-based security intrusion detectors on a centralized ground station, as well as on-board agents that may be deployed on a drone to detect anomalous behaviors of compromised drones of the fleet in real time.