Intelligent secure mobile edge computing for beyond 5G wireless networks

Computational task offloading at the mobile edge servers is a promising strategy to reduce latency and energy consumption in 5G wireless networks. In this paper, we study the problem of offloading decision and system design in the intelligent secure mobile edge computing (MEC) system with a UAV eavesdropper, where the eavesdropper can overhear the secure computational task from the user to the computational access point (CAP). The proposed framework is aimed to ensure the physical-layer security and decrease the latency and energy consumption in communication and computation. We firstly formulate the optimization of the MEC networks as a multi-objective optimization problem and then use a linear combination of the latency and energy consumption to measure the system cost. We devise an adaptive offloading strategy by incorporating the wireless bandwidth allocation and the transmit power allocation among users through the deep reinforcement learning (DRL) algorithm. In particular, we propose a deep Q-network (DQN) based strategy to automatically solve the optimization problem by designing the system state, action, and the reward function at detail. At last, we demonstrate the usefulness of the considered intelligent offloading strategy for the design of the MEC networks by comparing with the other schemes. The proposed strategy can perform significant system cost saving of the considered MEC networks. (C) 2021 Elsevier B.V. All rights reserved.