Pruned Causal Inference Enabled Active Deception for Covert Internet of Things (IoT)

In this paper, the problem of covert wireless communication in Internet of Things (IoT) systems under passive eavesdropping attacks is investigated. In the considered model, an IoT system intents to transmit a confidential signal from an origin to a destination under time and energy limitations. Meanwhile, a group of eavesdroppers attacks the IoT system by only passively listening to the confidential transmission without transmitting any signals. By detecting the oscillator leakage, the IoT devices can perceive the existence of nearby passive eavesdropping attacks. In this case, we propose an active deception scheme, with which the IoT devices will strategically transmit fake signals to distract and expose the passive eavesdroppers. Then, the passive eavesdroppers have to carefully choose the signal to intercept while avoiding exposure. The complex interaction among IoT devices and passive eavesdroppers is captured using tools from dynamic deception game, within which the conflicting goals of IoT system privacy and eavesdropping attacks are captured in multi-attribute utilities. Considering the challenges on searching the perfect bayesian equilibrium (PBE) in the partially observable, large-scale wireless IoT system, a pruned causal inference based approach is proposed. Simulation results show that the proposed method yields an up to 63.6% improvement in the convergence speed as compared to the traditional Bayesian-inference algorithm, while exhibiting an up to 229% improvement in security performance as compared to a traditional deception strategy that is designed by an open-loop algorithm.