基于敏感项集动态隐藏的用户隐私保护方法*

时空k-匿名由于适应移动性环境以及实现更为简单方便等特点，是当前LBS领域中被使用最广泛的模型。由于LBS在线及动态的特性，使传统的数据变形或重构方法不足以对抗从大量时空k-匿名数据集挖掘到的关联规则的用户隐私攻击。针对以上问题提出了基于敏感项集动态隐藏的用户隐私保护方法（SIDH）：感知敏感规则对应项集空间的正负边界，增量扩展原始快照查询匿名集数据，以敏感项集的动态隐藏净化敏感关联规则，最终实现用户隐私保护。通过对2612辆出租车的GPS数据生成的匿名集进行敏感项集隐藏实验，结果表明：SIDH方法隐藏敏感项集的数量和速度明显高于传统匿名方法，并且不会新增敏感项集。因此SIDH方法更能有效应对匿名集敏感关联规则的推理攻击，副作用较小。

User privacy protection method based on dynamic hiding of sensitive items

Recently, spatial-temporal k-anonymity has become a prominent approach among the field of LBS privacy protection. Analyzing numerous spatial-temporal k-anonymity datasets is beneficial for lots of LBS applications, but could cause an adversary to bring inference attacks which are not able to be handled by traditional methods of sensitive knowledge hiding. The reason is that traditional anti-attack methods can only solve the problems of privacy protection in offline and static environments, but do not meet privacy demands of online and dynamic LBS applications. To overcome these challenges, this paper presented a method called SIDH. The detailed procedure contained three phases: (a)offline mining frequent items of original anonymous datasets and computing the boundaries of items;(b)online perceiving the boundaries and avoiding negative boundaries based on the designed principles;(c)incrementally expanding snapshot anonymous sets and dynamically hiding sensitive items to purifying association rules. Finally, the conducted experiments on anonymous sets generated by GPS data of 2612 taxis demonstrate that SIDH can realize faster hiding of sensitive items than traditional methods ,and effectively deal with inference attacks on user privacy based on sensitive rules mined from anonymous datasets.