Affiliations: [a] School of Information Science and Engineering, Xiamen University, Xiamen, China | [b] School of Software, Xiamen University, Xiamen, China
Correspondence:
[*]
Corresponding author. Sin Hong Lin, School of Information Science and Engineering, Xiamen University, Xiamen, China. Tel.: +86 159 602 56029; Fax: +86 059 1833 72018; E-mail: eecs.lin@stu.xmu.edu.cn.
Abstract: Knowledge discovery on social network data can benefit general public, since these data contain latent social trends and valuable information. Recent research finds that preserving data privacy plays a vital role in knowledge discovery. Therefore, social network data need to be anonymized to preserve users’ identity before the data can be released for research purposes. In this paper, we model social network data as directed graphs with signed edge weights; formally define privacy, attack models for the anonymization problem. Based on our analysis, we develop a graph anonymization approach. The other main contribution is our graph clustering algorithm which can effectively group similar graph nodes into clusters with minimum cluster size constraints. Finally, we carry out a series of experiments to evaluate the effectiveness and utility of our approach on anonymizing social network data.
Keywords: Data publishing, K-anonymity, clustering algorithm, information system
