The interest in trajectory data has sensibly increased since the widespread of mobile devices. Simple clustering techniques allow the recognition of personal gazetteers, i.e., the set of main points of interest (also called stay points) of each user, together with the list of time instants of each visit. Due to their sensitiveness, personal gazetteers are usually anonymized, but their inherent unique patterns expose them to the risk of being de-anonymized. In particular, social trajectories (i.e., those obtained from social networks, which associate statuses and check-ins to spatial and temporal locations) can be leveraged by an adversary to de-anonymize personal gazetteers. In this paper, we propose DART as an innovative approach to effectively de-anonymize personal gazetteers through social trajectories, even in the absence of a temporal alignment between the two sources (i.e., they have been collected over different periods). DART relies on a big data implementation, guaranteeing the scalability to large volumes of data. We evaluate our approach on two real-world datasets and we compare it with recent state-of-the-art algorithms to verify its effectiveness.

自从移动设备的普及以来，对轨迹数据的兴趣已经明显增加。简单的聚类技术可以识别个人地名词典，即每个用户的主要兴趣点集（也称为停留点），以及每次访问的时间列表。由于其敏感性，个人地名词典通常被匿名化，但其固有的独特模式使他们面临被匿名化的风险。尤其是，攻击者可以利用社交轨迹（即从社交网络获得的，将状态和签到关联到空间和时间位置的轨迹）对个人地名词典进行匿名处理。在本文中，我们建议DART作为一种创新方法，以通过社交轨迹有效地使个人地名词典匿名化，即使两个源之间没有时间上的对齐（即，它们是在不同时期内收集的）。DART依赖于大数据实施，从而保证了对大量数据的可伸缩性。我们在两个真实的数据集上评估了我们的方法，并将其与最新的最新算法进行了比较，以验证其有效性。