连续数据发布的隐私保护研究
发布时间：2019-07-06 13:17
【摘要】：随着互联网的飞速发展,大数据时代已经到来,数据信息发布已成为很普遍的现象。大量的连续数据发布难免会泄露个人隐私信息,连续发布的隐私保护逐渐受到科技研究者的重视。发布后的数据具有较高的隐私保护度、较低的信息损失度和较好的可用性是数据发布隐私保护研究的重要目标。目前连续数据发布隐私保护研究还处于初级阶段,研究更有效的数据发布隐私保护算法迫在眉睫。首先分析了 LDMICA算法的优缺点,提出了一个静态数据集更新隐私保护算法—LDICA算法。LDICA算法使用LDMICA算法的聚类方式,利用方差计算每个属性的权重,再计算出每条记录的综合值大小。由综合值大小划分等价类,并合理的将剩余记录增加到等价类,使发布数据的每个等价类满足l-多样性以及具有相似的记录集。对LDICA算法进行了实验测试,LDICA算法不修改属性值,对划分后的等价类进行分割,并利用有损连接达到隐私保护,无信息损失度,多样性参数l选取总敏感属性值种类数的1/2具有最优的计算性能开销。接着结合静态数据发布的聚类思想以及置换匿名分割技术提出了一个适用于动态数据集更新连续发布的隐私保护算法—LDACA算法。由综合值大小划分等价类,实现数据的完全更新,使发布的数据表也相应的进行更新,并保持发布前的数据具有相同的签名。LDACA算法对数据的完全更新逐步处理删除记录模块、修改记录模块、伪记录表、增量表记录模块,处理增量表记录使用了 LDICA算法,使新增等价类满足l-多样性,同时保持原等价类签名不变。对LDACA算法进行了实验测试,LDACA算法的隐私泄露率相对于M-distinct算法降低了 20%,远远低于1/l,减小了信息损失度,可以有效的防止连接攻击。算法执行时间低,5s内执行完毕,性能较好,能有效的起到隐私保护的作用。
[Abstract]:With the rapid development of the Internet, the era of big data has arrived, data and information release has become a very common phenomenon. A large number of continuous data release will inevitably reveal personal privacy information, and the privacy protection of continuous release will be paid more and more attention by science and technology researchers. The published data have high privacy protection, low information loss and good availability are the important objectives of data release privacy protection research. At present, the research on privacy protection of continuous data release is still in its infancy, so it is urgent to study a more effective privacy protection algorithm for data release. Firstly, the advantages and disadvantages of LDMICA algorithm are analyzed, and a static dataset update privacy protection algorithm, LDICA algorithm, is proposed. LDICA algorithm uses the clustering method of LDMICA algorithm, uses variance to calculate the weight of each attribute, and then calculates the comprehensive value size of each record. The equivalence class is divided by the size of the synthesis value, and the remaining records are reasonably added to the equivalent class, so that each equivalent class of the published data satisfies l-diversity and has a similar record set. The LDICA algorithm is tested by experiments. The LDICA algorithm does not modify the attribute values, divides the divided equivalent classes, and uses lossy connections to achieve privacy protection, no information loss, and the diversity parameter l to select 1 鈮，
本文编号：2511039
[Abstract]:With the rapid development of the Internet, the era of big data has arrived, data and information release has become a very common phenomenon. A large number of continuous data release will inevitably reveal personal privacy information, and the privacy protection of continuous release will be paid more and more attention by science and technology researchers. The published data have high privacy protection, low information loss and good availability are the important objectives of data release privacy protection research. At present, the research on privacy protection of continuous data release is still in its infancy, so it is urgent to study a more effective privacy protection algorithm for data release. Firstly, the advantages and disadvantages of LDMICA algorithm are analyzed, and a static dataset update privacy protection algorithm, LDICA algorithm, is proposed. LDICA algorithm uses the clustering method of LDMICA algorithm, uses variance to calculate the weight of each attribute, and then calculates the comprehensive value size of each record. The equivalence class is divided by the size of the synthesis value, and the remaining records are reasonably added to the equivalent class, so that each equivalent class of the published data satisfies l-diversity and has a similar record set. The LDICA algorithm is tested by experiments. The LDICA algorithm does not modify the attribute values, divides the divided equivalent classes, and uses lossy connections to achieve privacy protection, no information loss, and the diversity parameter l to select 1 鈮，
本文编号：2511039
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2511039.html
