基于个性化k匿名隐私保护的资源推荐算法

Resource Recommendation Algorithm Based on K-anonymity for Generalizing User Query Requests

现有资源推荐算法依据用户喜好来进行资源推荐,未考虑用户主观并不愿意个人隐私或者兴趣爱好受到数据挖掘,也未考虑第三方服务器存在隐私泄露的危险。为解决用户的隐私保护问题,本文提出个性化k匿名隐私保护算法,将用户查询请求中的敏感属性经泛化后,构造逻辑上的匿名化查询请求等价类,采用数据轮转的方式使同一等价类中的用户相互随机转发接收到的其他用户的数据;由于每个人想要保护的隐私属性不同,其敏感属性的权重各不相同,故本文提出基于敏感属性的权重求和公式结合用户自主设置的敏感属性权重值,为平台用户推荐最优选择方案。安全性分析表明该方法有效抵御相似性攻击、背景知识攻击、俘获服务器攻击。实验表明,在牺牲一定时间效率的情况下该方法不仅满足匹配结果的正确性,且加强了资源推荐过程中隐私保护性能。

Existing resource recommendation algorithms make resource recommendation based on user preferences,without considering the subjective reluctance of users to personal privacy or the data mining of interests and hobbies,or the risk of privacy disclosure of third-party servers.To solve the problem of user privacy,this paper puts forward the K-anonymous privacy protection algorithm.After the generalization of sensitive attributes of user query requests,it constructs a logical anonymous query request equivalence class,uses the method of data rotary to enable users in the same equivalence class to randomly forward the received data from other users.Because the privacy attributes that each person wants to protect are different,the weight of its sensitive attribute is different,so this paper proposes the weight summation formula based on the sensitive attribute combined with the sensitive attribute weight value set by the user independently,and recommends the optimal selection scheme for the platform user.The security analysis shows that this method can effectively resist similarity attack,background knowledge attack and capture server attack.Experiments show that this method not only satisfies the correctness of matching results,but also enhances the privacy protection performance in the process of resource recommendation.