摘要：

针对目前移动边缘计算位置隐私保护算法导致用户任务卸载时服务质量降低的问题, 提出一种融合维诺图机制和ε-地理不可区分机制的位置隐私保护策略V-Geo。首先, 云服务器以地图中的边缘节点作为维诺图的生成元, 采用逐点插入法生成维诺图, 将地图划分为若干维诺格, 并将地图划分数据分发至地图中的用户; 其次, 用户以自身所在维诺格为安全区域, 融合ε-地理不可区分机制在该维诺格范围内基于自身真实位置生成一个虚假位置用于代替真实位置进行任务卸载; 最后提出一种融合时延、能耗和距离损失的服务质量损失模型, 该模型根据用户对时延、能耗和距离损失的不同需求评估用户任务卸载时的服务质量高低。以场景内所有用户的服务质量损失模型作为指标, 改变时延、能耗和距离损失的权重占比进行仿真, 仿真结果证明在同等的隐私保护程度下, V-Geo算法较基于本地差分隐私的改进算法(V-R)用户的服务质量损失平均减少了31.2%。同时证明了在不同用户数量和不同边缘节点数量下, V-Geo算法较其他算法依旧存在优势。

关键词: 移动边缘计算, 位置隐私保护, 差分隐私, 维诺图, 地理不可区分机制

Abstract:

Regarding the disadvantages of current Mobile Edge Computing(MEC) location privacy protection algorithms, which cause degradation of service quality when users offload tasks, a location privacy protection policy V-Geo integrating Voronoi diagram mechanism and ε-geo-indistinguishability mechanism is proposed. First, the cloud server takes the edge nodes in the map as the generating elements of the Voronoi diagram and produces the Voronoi diagram using the incremental insertion method. The map is divided into multiple Voronoi cells, and the map division data are distributed to users in the map. Second, users take the Voronoi cell where they are located as the secure area and integrate the ε-geo-indistinguishability mechanism to generate a fake location within the scope of this Voronoi cell based on their real location to replace the real location for task offloading. Finally, a service-quality loss model that integrates delay, energy consumption, and distance loss is proposed. This model evaluates service quality when users offload tasks according to different user requirements for delay, energy consumption, and distance loss. Using the service-quality loss model of all users in the scenario as an index, the weights of delay, energy consumption, and distance loss were varied for simulation. The simulation results prove that under the same privacy protection level, the average service quality loss of users of the V-Geo algorithm is reduced by 31.2% compared with the improved algorithm(V-R) based on local differential privacy. Moreover, under different numbers of users and edge nodes, the V-Geo algorithm maintains advantages over other algorithms.

Key words: Mobile Edge Computing(MEC), location privacy protection, differential privacy, Voronoi diagram, geo-indistinguishability