Secure reference-tracking with resource-constrained UAVs

In this paper we study a security problem for resource-constrained autonomous systems. We consider a UAV tasked with tracking a reference trajectory, and an attacker capable of compromising the measurements taken at certain sensors. We consider a probabilistic attack model, where the attacker executes denial of service attacks against a subset of sensors based on a Bernoulli process. We assume that sensors have different accuracy, reliability, and require different computational times to be activated. Our approach is based on the formalism of Markov Jump Linear Systems. We develop and numerically validate an optimal security countermeasure that probabilistically selects which sensor to use at different time instants, so as to balance performance and security, and ultimately minimize the UAV's expected tracking error.