Masking the Sensing Plan for Multifunction Radar in an Adversarial Setting

Multi-Function Radars (MFRs) employ a finite state controller to achieve the optimal cost of operation. The state-action transitions of the controller become subject to observation by potential adversaries. The adversaries' primary objective revolves around estimating the underlying state-action transition kernel and the sensing plan. How can the radar design a masking strategy against the adversary's pursuit of rule extraction? In this paper, we model the MFR's controller through a Markov Decision Process (MDP) framework, where state-action transitions are governed by the sensing plan. We demonstrate that the randomization of the sensing plan reduces the adversary's Fisher Information. Consequently, this reduction amplifies the variability in policy and transition kernel estimation errors, thwarting the adversary's accuracy in estimating the radar's sensing plan.