Poster: Towards Rule-Based Defense Mechanisms in Distributed UAV-Swarms

Due to increasing microchip capabilities and sinking hardware costs, the applications for embedded, distributed, real-time systems are becoming increasingly diverse. A common application sector is vehicles, such as cars, ships, airplanes, and UAVs. These systems often involve multiple layers of embedded components interacting across different levels of complexity. Due to the increasing complexity and openness of such systems, security measures are strongly required. Additionally, systems often encounter changing network topologies, e.g. in a UAV swarm, which incentivize the use of autonomous actions by agents whose connection to a central authority has been compromised. In this paper, we research a swarm of UAV drones that are subject to jamming attacks. We use lightweight Learning Classifier Systems (LCS) as a means to classify incoming attacks and/or connection losses and dynamically react based on mission objectives. The affected drones will be able to continue their (altered) mission objective, launch defensive countermeasures, and try to reestablish connection to other swarm members. The features are executed in the framework of the Artificial Hormone System (AHS) middleware. After simulation-based training and testing, the system is validated with real drones.