PCPIR-V:基于Spark的并行隐私保护近邻查询算法  

PCPIR-V:parallel privacy protected algorithms for nearest neighbor query based on Spark

作　　者：邓诗卓 姚继涛[1] 王波涛[1] 陈月梅[1] 袁野[1] 李艳辉[1] 王国仁[1] 

机构地区：[1]东北大学计算机科学与工程学院,辽宁沈阳110819

出　　处：《网络与信息安全学报》2016年第5期64-76,共13页Chinese Journal of Network and Information Security

基　　金：国家自然科学基金资助项目(No.61173030;No.61272181;No.61272182;No.61332014;No.61370154;No.61332006)~~

摘　　要：针对面向大数据的隐私保护查询效率低问题,利用CPIR保护程度高,实现了基于Spark的并行CPIR空间近邻查询隐私保护算法PCPIR-V,提出了基于Row和Bit的并行策略,同时提出并实现了基于聚类的PCPIR-V的缓存优化技术。利用均匀分布、高斯分布和真实数据对PCPIR-V进行了测试验证,在40个核心范围内,PCPIR-V具有良好的扩展性,PCPIR-V缓存优化技术计算时间与朴素PCPIR-V时间相比,平均减少了20%。To address the low-efficiency problem for query privacy protection on big data, parallel CPIR-V（PCPIR-V）, which had a high level of privacy protection for nearest neighbor query, was presented and implemented based on spark. Two parallel strategies for PCPIR-V, Row strategy and Bit strategy, were proposed. To avoid redundant multiplications, the repeated products were cached based on a clustering technique while computing CPIR on Spark. According to the evaluation results of PCPIR-V on three datasets, the scalablity of PCPIR-V is good until the number of core is larger than 40. The cost of PCPIR-V with the method of caching partial multiplication results is reduced by 20% averagely.

关 键 词：查询隐私保护 基于计算能力的隐私信息检索 SPARK 基于位置服务 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]