Abstract: It’s of very much importance that personal privacy information of social media users on the Internet should be protected. Prevailing research outcomes protecting privacy are mainly about general relational data, location based service and related trace data, and social network structure, while recognition of privacy exposure problems and the solutions solving them remain to be discovered fortemporal data of users’ behavior in social media. On the purpose of solving these problems, a multi-method and multi-option clustering based attacker model is proposed, and it’s capable of leaving social media users’ temporal patterns to much privacy exposure, which deserves considerable attention.

Key words: users&rsquo, behavior, temporal pattern, clustering, social media, privacy protection

摘要： 网络社交媒体用户个人隐私信息的保护具有极其重要的意义。现有的有关隐私保护的研究集中于一般关系型数据、位置和轨迹信息、社交网络关系等数据类型的隐私保护，而社交媒体数据结构的复杂性使得隐私泄露的情况变得更加隐蔽，现有针对其他类型数据的隐私泄露的判定方式和隐私保护方法难于直接用于社交媒体用户行为的时间模式分析。为了发现社交媒体复杂时间数据中存在的潜在隐私泄露问题，给出针对网络论坛用户的行为时间模式的隐私泄露挖掘方法，设计并实现了多方法、多选择的聚类攻击者模型，实验通过攻击社交媒体用户的行为模式数据集，发现了用户行为时间模式的特异性和网络论坛中广泛存在着的用户隐私泄露的问题，应当引起充分重视。

关键词: 用户行为, 时间模式, 聚类, 社交媒体, 隐私保护