中国邮电高校学报(英文) ›› 2014, Vol. 21 ›› Issue (2): 75-82.doi: 10.1016/S1005-8885(14)60289-2

• Others • 上一篇    下一篇

Background knowledge based privacy metric model for online social networks

成城1,张春红1,纪阳2   

  1. 1. 北京邮电大学
    2. 北京邮电大学电信工程学院
  • 收稿日期:2013-07-11 修回日期:2014-01-06 出版日期:2014-04-30 发布日期:2014-04-30
  • 通讯作者: 成城 E-mail:chengcheng20090901@gmail.com
  • 基金资助:

    基于Web的无线泛在业务环境体系架构、关键技术及示范应用研究

Background knowledge based privacy metric model for online social networks

  1. School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2013-07-11 Revised:2014-01-06 Online:2014-04-30 Published:2014-04-30
  • Contact: Cheng Cheng E-mail:chengcheng20090901@gmail.com
  • Supported by:

    the Social Network Based Cloud Service Technology for TV Content and Application (202BAH41F03).

摘要:

The data of online social network (OSN) is collected currently by the third party for various purposes. One of the problems in such practices is how to measure the privacy breach to assure users. The recent work on OSN privacy is mainly focus on privacy-preserving data publishing. However, the work on privacy metric is not systematic but mainly focus on the traditional datasets. Compared with the traditional datasets, the attribute types in OSN are more diverse and the tuple is relevant to each other. The retweet and comment make the graph character of OSN notably. Furthermore, the open application programming interfaces (APIs) and lower register barrier make OSN open environment, in which the background knowledge is more easily achieved by adversaries. This paper analyzes the background knowledge in OSN and discusses its characteristics in detail. Then a privacy metric model faces OSN background knowledge based on kernel regression is proposed. In particular, this model takes the joint attributes and link knowledge into consideration. The effect of different data distributions is discussed. The real world data set from weibo.com has been adopted. It is demonstrated that the privacy metric algorithm in this article is effective in OSN privacy evaluation. The prediction error is 30% lower than that of the work mentioned above

关键词:

privacy metric, social network, background knowledge, kernel regression

Abstract:

The data of online social network (OSN) is collected currently by the third party for various purposes. One of the problems in such practices is how to measure the privacy breach to assure users. The recent work on OSN privacy is mainly focus on privacy-preserving data publishing. However, the work on privacy metric is not systematic but mainly focus on the traditional datasets. Compared with the traditional datasets, the attribute types in OSN are more diverse and the tuple is relevant to each other. The retweet and comment make the graph character of OSN notably. Furthermore, the open application programming interfaces (APIs) and lower register barrier make OSN open environment, in which the background knowledge is more easily achieved by adversaries. This paper analyzes the background knowledge in OSN and discusses its characteristics in detail. Then a privacy metric model faces OSN background knowledge based on kernel regression is proposed. In particular, this model takes the joint attributes and link knowledge into consideration. The effect of different data distributions is discussed. The real world data set from weibo.com has been adopted. It is demonstrated that the privacy metric algorithm in this article is effective in OSN privacy evaluation. The prediction error is 30% lower than that of the work mentioned above

Key words:

privacy metric, social network, background knowledge, kernel regression