Trusted and Efficient Task Offloading in Vehicular Edge Computing Networks
Hongzhi Guo, Xiangshen Chen, Xiaoyi Zhou, Jiajia Liu
Abstract
To meet the computation-intensive and delay-sensitive requirements in smart driving, vehicular edge computing (VEC), which offloads the vehicles’ tasks to neighbor roadside units (RSUs) or other vehicles, is conceived as a promising approach. However, due to the untrustworthiness of nodes, there are still many security issues in VEC networks, exposing vehicles to severe risks. Recently, some researchers have explored trust evaluation mechanisms to filter out malicious attacks and ensure task vehicles’ security. Nevertheless, most of them only considered trustworthiness and ignored the offloading efficiency, and thus deploying them on VEC networks would bring high delay. Moreover, the accuracy of these trust evaluation works is fragile and unsatisfactory, considering some malicious attacks in VEC networks. To this end, we investigate the joint delay and trustworthiness optimisation problem for task offloading. Aiming to more accurately and stably assess vehicles’ trustworthiness in VEC networks, we first design a trust evaluation algorithm. After that, the joint optimization problem is defined as a Markov decision problem, and a deep reinforcement learning-based task processing method is developed, to reduce the task offloading delay in VEC networks. Extensive experiments verify that our solution has better performance in minimizing the offloading delay and enhancing the task processing trustworthiness.