摘要
查询日志的发布会泄露用户的隐私。提出一种基于差分隐私的查询日志匿名化算法:首先构建用户查询项模型进行相似度计算并利用所求结果对用户查询项模型进行聚类,其次在聚类过程中添加指数噪音来满足差分隐私,最后发布匿名化数据。实验表明:该算法有效地提高了查询日志的实用性和隐私保护程度。
A user profile set anonymization algorithm is proposed in this paper. Firstly, the user profile model is extracted to compute similarityand cluster with the computing result. Secondly, exponential noise is added in the cluster process for differential privacy. Finally, the anonymity data are released. Experimental results demonstrate that the proposed anonymization cluster algorithm promotesdata utility and level of privacy protection.
出处
《北京信息科技大学学报(自然科学版)》
2013年第5期24-27,31,共5页
Journal of Beijing Information Science and Technology University
基金
国家自然科学基金项目(61370139)
教育部人文社会科学项目(11YJC870011)
北京市教委科技计划面上项目(KM201211232014)
研究生教育提高项目(YJT201309)
关键词
差分隐私
隐私保护
查询日志
数据发布
匿名化
differential privacy
privacy protection
query log
data release
anonymization