Risk-based Security Measure Allocation Against Actuator Attacks

This article considers the problem of risk-optimal allocation security measures when actuators an uncertain control system are under attack. We consider adversary injecting false data into actuator channels. The attack impact is characterized by maximum performance loss caused a stealthy with bounded energy. Since random variable, due to uncertainty, we use Conditional Value-at-Risk (CVaR) characterize risk associated then allocating set minimize risk. assume that there only limited number available. Under this constraint, observe mixed-integer optimization problem. Thus relaxation techniques approximate Semi-Definite Program (SDP). also compare our method (i) across different measures: worst-case measure, average (nominal) and (ii) search algorithms: exhaustive greedy algorithms. depict efficacy approach through numerical examples.