基于Goldberg IT-PIR的最近邻LBS隐私查询协议研究及并行实现

发布时间：2018-05-13 05:27

本文选题：Goldberg信息论隐私信息检索协议 + LBS隐私保护　；参考：《西北农林科技大学》2017年硕士论文

【摘要】：随着移动互联网、社会网络以及大数据等新兴技术的快速发展,LBS隐私保护问题日益严重。与其他的LBS隐私保护技术相比,基于隐私信息检索(Private Information Retrieval,PIR)的LBS隐私保护技术具有隐私保护强度高,无需可信第三方,并且查询结果准确等优点。但现有的查询协议多假设LBS服务器是半诚实的,且存在抵御模式攻击而导致通信和计算代价较大、查询效率较低等的问题。针对这些问题,根据Goldberg的信息论隐私信息检索协议(Goldberg's Information-Theoretic Private Information Retrieval,Goldberg IT-PIR)计算复杂度较低,并且能够在一定条件下抵抗数据库服务器的恶意攻击的特点,论文提出了基于Goldberg IT-PIR的最近邻LBS隐私查询协议。主要研究工作及取得的成果如下:(1)提出了基于Goldberg IT-PIR的最近邻LBS隐私查询协议。针对因抵御模式攻击而导致通信和计算代价较大的问题,通过构建2层R树,使得不同用户的最近邻查询请求具有相同的PIR访问次数,而不需要添加无效的PIR检索来抵御模式攻击;对于查询效率较低的问题,应用Goldberg IT-PIR协议在线检索候选最近邻兴趣点,在LBS服务器可能是恶意的情况下(t个服务器相互串通、l-k个服务器崩溃、v个服务器返回错误的结果),能够保证用户安全地查询到正确的结果。(2)实现了基于Goldberg IT-PIR最近邻查询协议LBS服务器端查询响应的并行化。通过对LBS服务器端查询响应算法的分析,针对处理大规模查询请求时,查询效率较低的问题,引入并行化的Strassen矩阵乘法算法,利用多线程技术实现LBS服务器端查询响应的并行化计算。依据LBS隐私保护技术性能评估指标,通过实验分析算法的性能和准确度。实验结果表明,与Ghinita等人的最近邻查询算法相比,该算法的通信代价降低了大约25%,计算代价减少了大约87%;算法最近邻兴趣点的误差最大约为0.04%;当查询请求个数大于128时,并行化算法能够有效提高LBS服务器端的计算效率。
[Abstract]:With the rapid development of mobile Internet, social network and big data, privacy protection is becoming more and more serious. Compared with other LBS privacy protection techniques, the LBS privacy protection technology based on Private Information Retrieval PIR has the advantages of high privacy protection intensity, no need for trusted third parties, and accurate query results. However, most of the existing query protocols assume that the LBS server is semi-honest, and there are some problems such as high communication and computing cost, low query efficiency and so on. According to Goldberg's Information Theory Privacy Information Retrieval Protocol (Goldberger Information-Theoretic Private Information Retrieval Goldberg IT-PIR), the computational complexity is relatively low, and it can resist the malicious attack of the database server under certain conditions. In this paper, the nearest neighbor LBS privacy query protocol based on Goldberg IT-PIR is proposed. The main research work and the results obtained are as follows: (1) proposed the nearest neighbor LBS privacy query protocol based on Goldberg IT-PIR. Aiming at the problem that the communication and computation cost is high because of resisting the pattern attack, by constructing a two-layer R-tree, the nearest neighbor query requests of different users have the same number of PIR access times. It does not need to add invalid PIR retrieval to resist pattern attack. For the problem of low query efficiency, Goldberg IT-PIR protocol is applied to online search candidate nearest neighbor points of interest. In the case that LBS servers may be malicious, t servers collude with each other and 1 k servers crash and v servers return the wrong results, which can ensure that users can safely query the correct results. Parallelization of query response on the LBS server side of neighbor query protocol. Based on the analysis of query response algorithm on LBS server, the parallel Strassen matrix multiplication algorithm is introduced to solve the problem of low query efficiency when dealing with large scale query requests. The parallel computing of query response on LBS server is realized by multithreading technology. According to the performance evaluation index of LBS privacy protection technology, the performance and accuracy of the algorithm are analyzed experimentally. The experimental results show that compared with the nearest neighbor query algorithm proposed by Ghinita et al, the communication cost and computational cost of the algorithm are reduced by about 25 and 87, the maximum error of nearest neighbor point of interest is about 0.04, and when the number of query requests is more than 128, the maximum error of the nearest neighbor point of interest is about 0.04. Parallelization algorithm can effectively improve the computing efficiency of LBS server.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309

【相似文献】