NOMA-based secure computation offloading in marine Internet of things

Regarding the presence of multiple malicious eavesdropping device (ED) in the marine Internet of things (M-IoT), to ensure the secure computation offloading of unmanned surface vehicle (USV) to a high altitude platform (HAP), a non-orthogonal multiple access (NOMA) assisted transmission policy was employed, where a set of idle USV acted as jamming USV (JU) that interfered with the eavesdropping of the ED and formed NOMA clusters with the transmitting USV (TU). Considering the requirements such as energy constraints and security transmission, the offloading ratio, transmitting power, computation resource allocation and NOMA cluster selection were jointly optimized with the objective of minimizing the maximum task processing latency. To solve the proposed mixed-integer non-convex optimization problem, an algorithm combining deep deterministic policy gradient (DDPG) and cross-entropy was proposed. Simulation results show that the proposed algorithm can effectively reduce maximum task processing latency and ensure the security of the system.