基于差分隐私保护的兴趣点推荐算法设计

DESIGN OF POI RECOMMENDATION ALGORITHM BASED ON DIFFERENTIAL PRIVACY PROTECTION

兴趣点推荐算法收集用户的历史行为记录,根据收集到的记录推测用户偏好,结合用户偏好向用户推荐新的兴趣点。针对传统的兴趣点推荐过程中,用户的隐私信息容易被泄露的问题,利用差分隐私保护机制对用户信息进行保护,防止被恶意攻击。差分隐私保护实现机制主要包括指数机制和拉普拉斯机制,均被使用于地理位置隐私保护算法中。基于差分隐私保护的地理位置隐私保护算法根据数据集中各项记录的相互关系建立位置搜索树;运用指数机制并结合树的结构挑选出经常访问的k项纪录;对这k项记录添加拉普拉斯噪声,发布加噪后的位置搜索树。实验表明,该算法能在推荐效果不变的情况下,有效地保护用户的隐私信息。

Point of interest recommendation algorithm collects the user s history behavior records, infers the user s preference according to the collected record, and recommends the new point of interest to the user according to the user s preference. In the process of traditional interest point of recommendation, the privacy information of the user is easy to be leaked. In order to avoid this situation, the differential privacy protection mechanism was used to protect the user information from malicious attack. The implementation mechanisms of differential privacy protection mainly included exponential mechanism and Laplace mechanism, both of which were used in geographic privacy protection algorithms. The geographical location privacy protection algorithm based on differential privacy protection established a location search tree by the relationship between the records in the dataset. Then the frequently visited k-item records were selected by using the exponential mechanism and the structure of the tree. Finally, Laplace noise was added to the k-item records, and the location search tree after the noise was published. Experimental results show that the proposed algorithm can effectively protect users privacy information when the recommendation effect is invariant.