Physical-layer security based mobile edge computing for emerging cyber physical systems

This paper studies a secure mobile edge computing (MEC) for emerging cyber physical systems (CPS), where there exist K eavesdroppers in the network, which can threaten the task offloading. These K eavesdroppers can work either in a colluding mode where they cooperate to decode the secret message, or in a non-colluding mode where the eavesdroppers decode the message individually. For both eavesdropping nodes, we design the secure MEC system by devising a computation offloading ratio, transmit power and computational capability allocation to optimize the system performance mainly measured by the latency. In particular, a novel deep reinforcement learning (DRL) together with convex optimization (DRCO) is proposed, where the DRL is used to find a proper solution to the offloading ratio, while the convex optimization is implemented to solve the allocation of transmission power and computational capability. Simulation results show that the proposed DRCO method is superior to other conventional methods, and can provide a guaranteed secrecy and latency.