Efficient privacy-preserving content recommendation for online social communities
Efficient privacy-preserving content recommendation for online social communities
复制标题
在线社交社区的高效隐私保护内容推荐
DOI：
10.1016/j.neucom.2016.09.059
复制
发表时间：
2017-01-05
期刊：
影响因子：
6
通讯作者：
Gu, Ning
中科院分区：
文献类型：
--
作者：
Li, Dongsheng;Lv, Qin;Gu, Ning
在网络社交社区中，许多推荐系统使用协同过滤，这是一种基于具有相似兴趣的其他用户所喜欢的内容进行推荐的方法。在此过程中会出现隐私问题，因为敏感的个人信息（例如内容兴趣）可能会被收集并透露给推荐服务器。现有的隐私保护协同过滤技术要么为了隐私牺牲效率，要么牺牲准确性，这对于拥有大量用户的网络社交社区是不合适的。在本文中，我们提出了YANA（“你并不孤单”的缩写），一种针对网络社交社区用户的基于用户群组的隐私保护推荐系统。在该系统中，用户被组织成具有不同兴趣的群组，并通过特定兴趣的伪用户与推荐服务器进行交互，这样单个用户的个人兴趣信息对服务器是隐藏的。提出了一套安全的多方计算协议和推荐策略，以在推荐过程中保护用户免受群组成员的隐私侵犯。在移动设备和台式计算机上都实现了一个原型系统，使用真实世界数据进行的评估表明，YANA能够有效保护用户隐私，同时实现高推荐质量和能源效率。
In online social communities, many recommender systems use collaborative filtering, a method that makes recommendations based on what are liked by other users with similar interests. Privacy issues arise in this process, as sensitive personal information (e.g., content interests) may be collected and disclosed to the recommender server. Existing privacy-preserving collaborative filtering techniques trade either efficiency or accuracy for privacy, which are not suitable for online social communities with large amount of users. In this paper, we propose YANA (short for "you are not alone"), a user group-based privacy-preserving recommender system for users in online social communities. In this system, users are organized into groups with diverse interests and interact with the recommender server via interest-specific pseudo users, so that individual user's personal interest information remains hidden from the server. A suit of secure multi-party computation protocols and recommendation strategies are proposed to protect user privacy from group members in the recommendation process. A prototype system has been implemented on both mobile devices and desktop computers, and evaluation using real-world data demonstrates that YANA can effectively protect users' privacy, while achieving high recommendation quality and energy efficiency.