loading page

Detecting Changes in Strategy of a Cognitive Radar: a Stochastic Revealed Preference Framework
  • Anoop C V,
  • Anup Aprem
Anoop C V

Corresponding Author:[email protected]

Author Profile
Anup Aprem
National Institute of Technology, National Institute of Technology Calicut


In this paper, we introduce a framework for detecting changes in strategy of an adversarial cognitive radar in the inverse learning context. We model the cognitive radar as a constrained utility maximizer and formulate the problem of detection of change in strategy as a changepoint detection problem in the revealed preference setting. We address the revealed preference strategy changepoint detection problem in a Bayesian setting, wherein we model the utility of cognitive radar as a random direction vector, that follows the von Mises-Fisher distribution with unknown parameters. The main contribution of the paper is the development of Bayesian changepoint detection algorithm that detects the changes in strategy of an adversarial cognitive radar, in an inverse learning context under stochastic revealed preference framework. The main advantages of the proposed method are (i) The proposed method uses a Hamiltonian Monte Carlo sampling algorithm that exploits the Afriat's theorem as well as a subsetting structure that arises in the posterior, and hence is devoid of the computational burden of solving optimization problems in existing techniques. (ii) Numerical results demonstrate the ability to determine the changes in the beam allocation strategy of a cognitive radar in noise free and noisy adversarial settings. The proposed approach inherently resists observation noise compared to the existing naive approaches. (iii) The proposed changepoint detection performs, on an average, five times faster compared to a naive, Generalized Likelihood Ratio approach for changepoint detection. Further, we also demonstrate that the proposed method generalizes existing changepoint detection under revealed preference setting and can be used for applications such as social media and finance. The changepoint detection framework, in the inverse learning context, is useful in the design of both electronic counter measures and electronic counter-counter measures.
15 May 2024Submitted to TechRxiv
20 May 2024Published in TechRxiv