面向Mashup多敏感属性数据集的隐私保护方法研究

发布时间：2019-03-01 19:01

【摘要】：Mashup是目前Internet上广受关注的基于Web的数据集成应用。作为一种常见的数据聚合应用,Mashup为数据的交换和共享提供了强有利的支持。数据聚合发布时会涉及多个数据发布单位,这些数据源间的连接操作往往会导致严重的隐私泄露问题,也极易产生敏感信息在数据发布单位之间泄露的情况。同时,从众多数据源聚合而来的数据必然包含大量属性,对高维度数据匿名化后极易造成数据过度失真的情况。数据聚合发布下的隐私保护是一个重要且富有挑战的难题。PHDMashup算法是针对数据聚合发布中的隐私保护问题而提出,它采用LKC-Privacy保护模型,结合自顶向下特化的方法,实现了数据聚合发布时的隐私保护。但是在数据聚合的过程中涉及众多数据提供方,需要进行匿名化处理的属性数量必然是巨大的,PHDMashup算法要求对所有属性构建的泛化树的全部有效节点进行特化,不仅造成了时间和空间的浪费,也带来了繁重的计算量。本文在此基础上对PHDMashup算法进行改进,提出了 NPHDMashup算法,通过减少特化节点的方式提高了算法的执行效率。另外,针对以上两种算法中由于数据提供方之间大量的信息交流造成时间耗费的情况,提出了一种改进型数据聚合隐私保护算法SPHDMashup,通过引入Server作为中间件,数据提供方直接与Server进行信息交流,不仅大大提高了算法效率,也在很大程度上减少了数据提供方的工作量。而且,以Mashup模式汇聚的资源具有多源、异质、结构复杂的特点,异构问题会影响数据提供方之间共享属性的数据处理,本文提出通过构建映射表的方法实现局部数据模型向公共模型的转换,解决数据聚合时的语义异构问题。最后通过实验对提出的算法进行评估,通过与原算法进行分析比对,验证了算法的优越性,并对算法中存在的不足进行了分析,讨论了今后算法的改进方向。
[Abstract]:Mashup is a widely concerned Web-based data integration application on Internet. As a common data aggregation application, Mashup provides strong support for data exchange and sharing. Data aggregation and publishing will involve multiple data publishing units, the connection between these data sources will often lead to serious privacy disclosure problems, but also very easy to generate sensitive information leakage between data publishing units. At the same time, the data aggregated from many data sources must contain a large number of attributes, which can easily result in over-distortion of data after anonymization of high-dimensional data. Privacy protection under data aggregation publishing is an important and challenging problem. PHD Mashup algorithm is proposed to solve the privacy protection problem in data aggregation publishing. It adopts LKC-Privacy protection model and combines top-down specialization method. The privacy protection of data aggregation and publishing is realized. However, many data providers are involved in the process of data aggregation, and the number of attributes that need to be anonymized must be huge. The PHDMashup algorithm requires the specialization of all effective nodes of the generalization tree constructed by all the attributes. It not only causes waste of time and space, but also brings heavy calculation. In this paper, the PHDMashup algorithm is improved and the NPHDMashup algorithm is proposed, which improves the efficiency of the algorithm by reducing the specialized nodes. In addition, an improved data aggregation privacy protection algorithm, SPHDMashup, which uses Server as a middleware, is proposed to solve the problem of time-consuming caused by a large amount of information exchange between data providers in the above two algorithms. The direct exchange of information between the data provider and Server not only improves the efficiency of the algorithm, but also reduces the workload of the data provider to a great extent. Moreover, the resources converged in Mashup mode have the characteristics of multi-source, heterogeneous and complex structure, and heterogeneous problems will affect the data processing of shared attributes among data providers. In this paper, a mapping table is proposed to realize the transformation from local data model to common model, so as to solve the problem of semantic heterogeneity in data aggregation. Finally, the proposed algorithm is evaluated by experiments, compared with the original algorithm, the superiority of the algorithm is verified, the shortcomings of the algorithm are analyzed, and the improvement direction of the algorithm in the future is discussed.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP309

【参考文献】