Optimal transmission scheduling over multihop networks under eavesdropping-DoS mixed attacks

This paper investigates the optimal transmission strategy for remote state estimate over signal-to-interference-plus-noise ratio (SINR)-based multihop network, against against an eavesdropping-DoS mixed attack, which is able to use the active eavesdropping information to jam the network. An intelligent sensor is used to transmit the local state estimate to a remote estimator (RE). To save energy, the multihop network is deployed to relay data packets from the sensor to the RE. To minimize the energy consumption and the known estimation error covariance at the RE, while maximizing the unknown eavesdropper error covariance, the transmission scheduling is formulated as a modified Markov decision process (MDP) by introducing a belief state probability distribution and incorporating historical actions into the state. A Clipped Double Q-learning algorithm is designed to learn the approximate optimal strategy online. Numerical simulation example is provided to illustrate the developed results.