期刊导航

论文摘要

基于图论的边缘计算信任评估优化模型

Optimization Scheme of Trust Model Based on Graph Theory for Edge Computing

作者:杜瑞忠(河北大学 网络空间安全与计算机学院, 河北 保定 071002);许琨琪(河北大学 网络空间安全与计算机学院, 河北 保定 071002);田俊峰(河北大学 河北省高可信信息系统重点实验室, 河北 保定 071002)

Author:DU Ruizhong(School of Cyberspace Security and Computer, Hebei Univ., Baoding 071002, China);XU Kunqi(School of Cyberspace Security and Computer, Hebei Univ., Baoding 071002, China);TIAN Junfeng(Key Lab. on High Trusted Info. System in Hebei Province, Hebei Univ., Baoding 071002, China)

收稿日期:2019-09-16          年卷(期)页码:2020,52(3):150-158

期刊名称:工程科学与技术

Journal Name:Advanced Engineering Sciences

关键字:边缘计算;信任模型;信任路径;冗余优化

Key words:edge computing;trust model;trust path;redundancy optimization

基金项目:国家自然科学基金项目(61572170;61170254);河北省自然科学基金项目(F2018201153);河北省高等学校科学技术研究基金项目(ZD2016043);河北省物联网数据采集与处理工程技术研究中心基金项目(河北065201)

中文摘要

针对边缘计算环境中的设备资源受限、现有信任模型忽略计算负载与信任路径冗余的问题,提出了一种基于图论的边缘计算信任评估优化模型。首先,基于边缘计算构建信任模型的体系架构,将边缘设备间复杂庞大的信任关系抽象成有向加权图,并对设备间的信任关系进行定义说明,再采用基于信息熵理论的自适应聚合方法对信任值进行聚合计算,修正多源信任之间的差异度;其次,通过添加信任阈值、路径长度限制、滑动窗口等多重约束条件,事先过滤明显不符合信任要求的节点和信任边,降低不必要的计算消耗;最后,利用改进后的深度优先搜索算法(depth first search, DFS),在信任路径搜索过程中规避冗余信任边,从而避免环路以及节点绕路问题,并采用递归函数Combine聚合反馈信任值。使用MATLAB仿真软件确定实验参数,验证模型区分恶意节点与正常节点的能力。并在交互成功率、时间开销以及能量开销3个方面进行实验,将本文模型与PSM模型、RFSN模型以及随机选择模型进行对比。实验结果表明,相较于其他模型,本文模型在不同诚实程度的网络环境下都能快速达到稳定状态,且时间与能量开销均低于其他模型,证明该模型在保证有效性的同时,能够在一定程度上减轻边缘设备的资源开销,提高网络的生存周期。

英文摘要

In order to solve the problems of the limited devices resources in the edge computing environment and the negligence of computing load as well as the redundant trust path in the existing trust models, a trust evaluation optimization model of the edge computing was proposed based on the graph theory. First, an architecture of trust model based on edge computing was built, of which the complex and huge trust relationship between edge devices was abstracted into a directed weighted graph, and the trust relationship between devices was defined and explained. Then, an adaptive aggregation method based on the information entropy theory was used to aggregate the trust value, which could correct the difference between multi-source trust. Secondly, the constraints of trust threshold, path length restriction and sliding window were added. With these multiple constraints, the nodes and trust edges that obviously do not meet the trust requirements were filtered in advance, which reduces unnecessary computing consumption. Finally, an improved depth first search algorithm was used to filter redundant trust edges, which could avoid loop and node detour problems in the trust path search process. The recursive function Combine was further used to aggregate the feedback trust value. The MATLAB simulation software was used to determine the experiment parameters, and verified the model’s ability of distinguishing malicious nodes from normal nodes. The proposed model was compared with PSM model, RFSN model and random selection model in interaction success rate, time cost and energy cost. The experimental results show that compared with other models, the proposed model could achieve stable state quickly in network environments with different honesty degrees, and the time and energy costs are lower than that of other models. The proposed model can reduce the resource overhead of edge devices and improve the network life cycle while ensuring the effectiveness.

关闭

Copyright © 2020四川大学期刊社 版权所有.

地址:成都市一环路南一段24号

邮编:610065