基于差分隐私的不确定数据频繁项集挖掘算法

Frequent itemsets mining for uncertain data based on differential privacy

基于不确定数据的频繁项集挖掘算法已经得到了广泛的研究。对于记录用户敏感信息的不确定数据,攻击者可以利用自己掌握的背景信息,通过分析基于不确定数据的频繁项集从而获得用户的敏感信息。为了从不确定的数据集中挖掘出基于期望支持度的前K个最频繁的频繁项集,并且保证挖掘结果满足差分隐私,提出了FIMUDDP(frequent itemsets mining for uncertain data based on differential privacy)算法。FIMUDDP算法利用差分隐私的指数机制和拉普拉斯机制确保从不确定数据中挖掘出的基于期望支持度的前K个最频繁的频繁项集和这些频繁项集的期望支持度满足差分隐私。通过对FIMUDDP进行理论分析和实验评估,验证了FIMUDDP算法的有效性。

Frequent itemsets mining for uncertain data has been studied extensively. For uncertain data which recorded private information of users,an attacker made use of background knowledge to obtain private information of users by analyzing frequent itemsets mined from uncertain data. In this regard,this paper proposed a new algorithm,denoted as FIMUDDP（ frequent itemsets mining for uncertain data based on differential privacy）,to mine the top K most frequent itemsets based on expected support from uncertain data and satisfy differential privacy. FIMUDDP algorithm applied exponential mechanism and Laplace mechanism in differential privacy to ensure differential privacy for the top K most frequent itemsets based on expected supports of these frequent itemsets respectively. Finally,through analyzing FIMUDDP from theory and experiment evaluation,the results demonstrate the effectiveness of FIMUDDP.