Secure Task Offloading and Resource Scheduling in Maritime Edge Computing Systems

Mobile edge computing (MEC) scheme can enable maritime Internet of things (MIoT) to support computing-intensive and delay-sensitive maritime applications. However, due to the broadcast characteristics of wireless signals, the information in the legitimate link may be eavesdropped by eavesdroppers. This paper considers a maritime edge computing system with an eavesdropper where unmanned surface vehicles (USVs) and an unmanned aerial vehicle (UAV) cooperate to offload and compute tasks. The task offloading and resource scheduling problem is modeled as a Markov decision process (MDP), in which the objective is to maximize the overall performance of the system with the consideration of the physical layer security (PLS). An offloading decision is made based on the channel qualities between USVs and UAV, task queue status and task information of USVs. In order to deal with the high-dimensional curse of state and action space, we propose a computational offloading algorithm based on the twin delayed deep deterministic policy gradient (TD3) to learn the optimal strategy without requiring a priori knowledge of the environment. Numerical experiments show that compared with other benchmark algorithms, our proposed learning algorithm has better overall system performance while ensuring secure communication.