支持个性化和协同标记质量的隐私保护-hku.pdfVIP

支持个性化和协同标记质量的隐私保护 泰山学堂2012 级计算机科学与技术专业 李晓东 指导教师 孙宇清 摘 要：协同标记平台正面临着用户隐私泄露的风险，而支持个性化和协同标记质量的隐私保护可以 在不影响协同标记质量的前提下大幅度提高协同标记平台的安全性。本文首先分析了协同标记平台隐私安 全的研究现状，并实现了协同标记身份识别系统STBUI ，它集成了协同标记平台的用户情感特征提取算法， 用户特征概率分布的分层提取算法，集成资源信息的层次特征提取算法，以及时空、关联度等环境特征提 取算法。与相关工作对比发现，STBUIS在识别精度上有很大的进步。基于身份识别系统中发现的安全漏洞， 本文设计了协同标记用户隐私保护系统CTUPPS，它同时关注了协同标记平台上应对身份攻击的隐私保护 能力和隐私保护代价问题两个方面。在对已有协同标记隐私保护系统进行身份识别攻击后发现，CTUPPS 可以在对原有协同推荐系统产生更小影响的前提下，保证协同标记系统的隐私保护能力。 关键词：协同标记；隐私保护；身份识别；网络标签；特征提取 Abstract: Collaborative tagging is faced with the risk of user privacy disclosure, and privacy protection for collaborative tagging quality and user personalization is required to improve the security without significantly affecting collaborative tagging quality. Firstly, this paper analyzes the research status of collaborative tagging privacy protection problem. Then, social tags based user identification system STBUIS is achieved, which integrates four algorithms, including user emotional feature extraction algorithm, layered characteristics probability distribution extraction algorithm, resource information integrated layered feature extraction algorithm, and objective feature extraction algorithm, such as time, location and user relationship. Compared with the related works, STBUIS has a higher precision. Based on these security flaws, the collaborative tagging user privacy protection system CTUPPS is achieved. CTUPPS not only pays attention to protect users from identification attacking, but also tries to decrease the costs of privacy protection. Compared with the existing collaborative tagging privacy protection systems, CTUPPS can keep the privacy level with smaller impaction on the collaborative recommendation system w