The vast majority of today's critical infrastructure is supported by numerousfeedback control loops and an attack on these control loops can have disastrousconsequences. This is a major concern since modern control systems are becominglarge and decentralized and thus more vulnerable to attacks. This paper isconcerned with the estimation and control of linear systems when some of thesensors or actuators are corrupted by an attacker. In the first part we look atthe estimation problem where we characterize the resilience of a system toattacks and study the possibility of increasing its resilience by a change ofparameters. We then propose an efficient algorithm to estimate the statedespite the attacks and we characterize its performance. Our approach isinspired from the areas of error-correction over the reals and compressedsensing. In the second part we consider the problem of designingoutput-feedback controllers that stabilize the system despite attacks. We showthat a principle of separation between estimation and control holds and thatthe design of resilient output feedback controllers can be reduced to thedesign of resilient state estimators.
展开▼