Privacy-by-Design Distributed Offloading for Vehicular Edge Computing
Privacy-by-Design Distributed Offloading for Vehicular Edge Computing
复制标题
DOI：
10.1145/3344341.3368804
复制
发表时间：
2019-12
期刊：
影响因子：
--
通讯作者：
Weibin Ma;Lena Mashayekhy
中科院分区：
文献类型：
--
作者：
Weibin Ma;Lena Mashayekhy
车辆边缘计算（VEC）是一种分布式计算范式，利用智能车辆（SV）作为计算云（边缘节点），凭借其固有的属性，如移动性、低运营成本、灵活部署和无线通信能力。 VEC通过扩大计算覆盖范围并进一步提高设备的服务质量（QoS）来扩展边缘计算服务。由于车载能量和 SV 安装云的计算能力有限，单个车辆可能无法执行大量任务并保证其所需的 QoS。为了解决这个问题，超载的车辆可以通过将其任务卸载到其他可用的联网车辆来完成其巨大的工作量。然而，数据隐私和可访问性至关重要，需要在卸载时予以考虑。在本文中，我们提出了 VEC 的隐私设计卸载解决方案，以满足用户需求的延迟要求并降低车辆的能耗。我们将数据保护卸载问题（DROP）制定为整数程序并证明其 NP 难度。为了提供计算上易于处理的解决方案，我们利用图论提出了三种分布式算法来解决这个问题。我们通过大量实验评估了我们提出的算法的性能，并将它们与 IBM ILOG CPLEX 获得的最佳结果进行比较。结果证明了我们提出的算法在提供实用的隐私设计卸载解决方案方面的灵活性、可扩展性和成本效率，从而实现了云到物连续体的边缘服务。
Vehicular Edge Computing (VEC) is a distributed computing paradigm that utilizes smart vehicles (SVs) as computational cloudlets (edge nodes) by virtue of their inherent attributes such as mobility, low operating costs, flexible deployment, and wireless communication ability. VEC extends edge computing services by expanding computing coverage and further improving quality-of-services (QoS) for devices. Due to limited onboard energy and computation capabilities of SV-mounted cloudlets, a single vehicle might not be able to execute a large number of tasks and guarantee their desired QoS. To address this problem, the overloaded vehicle can fulfill its overwhelming workload by offloading its tasks to other available connected vehicles. However, data privacy and accessibility are of critical importance that need to be considered for offloading. In this paper, we propose privacy-by-design offloading solutions for VEC to facilitate latency requirements of user demands and reduce energy consumption of vehicles.We formulate the Data pRotection Offloading Problem (DROP) as an Integer Program and prove its NP-hardness. To provide computationally tractable solutions, we propose three distributed algorithms by leveraging graph theory to solve this problem. We evaluate the performance of our proposed algorithms by extensive experiments and compare them to the optimal results obtained by IBM ILOG CPLEX. The results demonstrate the flexibility, scalability, and cost efficiency of our proposed algorithms in providing practical privacy-by-design offloading solutions enabling edge services along the cloud-to-thing continuum.