Secure Routing in Multihop Ad-Hoc Networks with SRR-based Reinforcement Learning

In this letter, a reinforcement learning-assisted secure routing methodology is proposed for multihop ad-hoc networks in the presence of multiple eavesdroppers. Specifically, secure relay region (SRR) is firstly proposed, which depicts the distribution of the relays forwarding the information securely. Moreover, a SRR-based on-policy Monte Carlo methodology is derived, aiming at accelerating the convergence of routing. The secrecy connection probability is also calculated, which indicates the secure performance of different routes. Simulation results show that our proposed SRR-based reinforcement learning methodology can select the secure route efficiently and fast, which is also robust to the time-varying available relays.