###
工程科学与技术:2019,51(2):137-143
←前一篇   |   后一篇→
本文二维码信息
码上扫一扫!
基于局部差分隐私的电动汽车充电位置隐私汇聚
(1.武汉大学 计算机学院, 湖北 武汉 430072;2.湖北省水利水电科学研究院, 湖北 武汉 430070;3.中南民族大学 计算机学院, 湖北 武汉 430074)
Private Electric Vehicle Charging Location Aggregation Based on Local Differential Privacy
(1.School of Computer Sci., Wuhan Univ., Wuhan 430072, China;2.Hubei Water Resources Research Inst., Wuhan 430070, China;3.College of Computer Sci., South-Central Univ. for Nationalities, Wuhan 430074, China)
摘要
图/表
参考文献
相似文献
本文已被:浏览 2076次   下载 1118
投稿时间:2018-09-24    修订日期:2019-03-04
中文摘要: 电动汽车频繁接入充电桩充电而产生的位置数据对优化充电桩布置、指导电力调度具有重要意义。然而充电位置数据对于汽车用户来说属于隐私信息。为防止汽车用户的隐私泄露,亟需探索研究隐私汇聚充电位置数据的方法。采用局部差分隐私技术保护电动汽车充电位置数据,通过引入贝叶斯随机多伪隐私算法设计一种基于分区的隐私保护充电位置数据汇聚方法。该方法利用贝叶斯随机多伪隐私算法设计了一个用于本地化扰动充电位置数据的局部混淆算法,然后,结合随机多伪算法的重构算法设计了满足稀疏、样本量小等特点的充电位置数据的隐私汇聚方法。同时,在保证隐私保护水平的前提下,通过对位置域进行划分以缩小隐私位置域,进一步提高汇聚结果的可用性。对所设计方法的隐私性进行分析。最后,在正态分布、均匀分布、峰值分布和随机分布4种不同的合成数据集以及公开的Gowalla数据集上进行验证。实验结果表明:在相同隐私水平的条件下,所设计的方法在可用性方面优于基于随机映射矩阵的隐私汇聚方法。
Abstract:The charging location data generated by electric vehicles frequently accessing charging piles for charging are of great significance for optimizing the arrangement of charging piles and guiding the electric power dispatching. However, charging location data are private information for vehicle users. In order to prevent the leakage of the privacy of these users, it is urgent to explore a way of private charging location data aggregation. Therefore, a local differential privacy technology is adopted to preserve the charging location data of electric vehicles. A partition-based privacy preservation charging location data aggregation method is proposed by introducing Bayesian randomized multiple dummies algorithm. The method employs the Bayesian randomized multiple dummies algorithm to design a local obfuscation algorithm for locally perturbing a vehicle's charging location. Then, the private location aggregation method for charging location data with the characteristics of sparseness and small size samples is designed by combining reconstruction algorithm of the randomized multiple dummies algorithm. At the same time, under the premise of ensuring the level of privacy preservation, the whole location domain is divided to narrow the privacy location domain, thereby further improving the utility of aggregation result. The privacy analysis of the proposed method is given. Finally, experimental results on four different synthetic datasets, namely, uniform distribution, normal distribution, peak distribution and random distribution, as well as the public Gowalla dataset are carried out. The experimental results show that the proposed method is superior to the existing randomized projection matrix based private aggregation method in terms of utility under the same privacy level.
文章编号:201801051     中图分类号:    文献标志码:
基金项目:国家自然科学基金资助项目(61872431);湖北省技术创新专项资助(2018AAA046);武汉市应用基础研究计划资助项目(2017060201010162)
作者简介:熊星星(1989-),男,博士生.研究方向:隐私保护、机器学习.E-mail:xiong_xx@whu.edu.cn
引用文本:
熊星星,刘树波,李丹,李永凯,王俊.基于局部差分隐私的电动汽车充电位置隐私汇聚[J].工程科学与技术,2019,51(2):137-143.
XIONG Xingxing,LIU Shubo,LI Dan,LI Yongkai,WANG Jun.Private Electric Vehicle Charging Location Aggregation Based on Local Differential Privacy[J].Advanced Engineering Sciences,2019,51(2):137-143.