基于ELM集成和半监督聚类的SNS隐私保护

针对各类网络数据中存在着大量的无标记数据,导致了SNS(social network service)隐私保护中数据可用性相对较差的问题,本文提出一种基于Bagging的ELM(extreme learning machine)集成算法,并将其与基于Seeds集的半监督聚类算法相结合应用于隐私保护.该算法首先利用ELM-Bagging集成方法对无标记数据进行标记,并将新标记的数据加入Seeds集以扩大其规模,然后采用基于Seeds集的半监督聚类实现K-匿名.实验结果表明,该算法在有效保护隐私的同时,提高了发布数据的可用性.

SNS privacy protection based on the ELM integration and semi-supervised clustering

Because of vast number of unlabeled data and small amount of labeled data in SNS,the availability of data is relatively poor for the study of SNS privacy preservation.In order to solve the above problem,this paper propose an ELM ensemble algorithm based on Bagging combined with semi-supervised clustering,based on Seeds set for privacy preserving.The main process is as follows: firstly,the ensemble ELM is used to label the unlabeled data to enlarge the scale of Seeds set;secondly,the Seeds set is used to initialize the center of clustering;and finally,the algorithm adopts semi-supervised clustering to achieve K-anonymization.Experimental results show that the method can improve the usability of the released data while preserving privacy.