Physical Layer Security-assisted Partial Computation Offloading in Mobile Edge Computing

With mobile edge computing (MEC), resource-constrained Internet of Things (IoT) users can offload computation tasks to edge servers in proximity for processing, which can reduce the service latency and energy consumption. Due to the broadcast nature of wireless communications, sensitive information can be leaked during computation offloading, in presence of adversaries or eavesdroppers. This work aims to improve the latency, energy efficiency and security in computation offloading in dynamic MEC environments. Specifically, partial offloading strategy is considered and a physical layer security-assisted scheme is developed to achieve multiple objectives, including maximizing the number of completed tasks before their respective deadlines and minimizing energy consumption, while providing enhanced security in a long run. A Markov decision process (MDP) with discrete-continuous action spaces is formulated and a deep reinforcement learning (DRL) method named hybrid-action deep deterministic policy gradient (HDDPG) is proposed to optimize the offloading ratio, friendly jammer selection, and computing power allocation. Simulation results demonstrate that the proposed HDDPG outperform the state-of-art deep DRL baselines in terms of the number of completed tasks before deadlines and energy costs while satisfying the security requirements.