Synthesis of the Supremal Covert Attacker Against Unknown Supervisors by Using Observations

In this paper, we consider the problem of synthesizing supremal covert damage-reachable attacker, in setup where model supervisor is unknown to adversary but has recorded a (prefix-closed) finite set observations runs closed-loop system. The synthesized attacker needs ensure both damage-reachability and covertness against all supervisors which are consistent with given observations. There gap between de facto supremality, assuming known, supremality that can be attained limited knowledge supervisor, from adversary's point view. We exercise sensor replacement/deletion attacks actuator enablement/disablement attacks. solution methodology proposed work reduce synthesis plant observations, safe for certain transformed plant, shows decidability observation-assisted problem. effectiveness our approach illustrated on water tank example adapted literature.