Abstract The rise of mobile computing has generated huge amount of trajectory data. Since these data are valuable for many people, publishing them while providing adequate individual privacy protection has been a challenging task. In this paper, we present an algorithm for protecting sensitive place visits in privacy-preserving trajectory publishing. By generalizing sensitive places using sensitive zones, and distorting the sub-trajectories within the sensitive zones based on differential privacy, our method not only prevents leakage of sensitive place visits, but also preserves individual movement information. It contains two critical components. First, we generate sensitive zones around sensitive places based on human mobility patterns and the mobility model. The sensitive zones are formed in such a way that the adversary background knowledge does not increase the adversary's belief in whether the trajectory has stopped at a sensitive place or not. Second, to prevent excessive individual movement information loss and sensitive place visit leakage within the sensitive zones, we select reliable segments from the sub-trajectories therein, model the reliable segments as an exploration tree, and synthesize the ɛ– differentially-private sub-trajectories. Our experiments on a real-world dataset show that our method provides good utility, and our sub-trajectory synthesis method preserves detailed information of individual movements.

中文翻译：

---

在隐私保护轨迹发布中保护敏感的地方访问

摘要 移动计算的兴起产生了海量的轨迹数据。由于这些数据对许多人来说都很有价值，因此在提供足够的个人隐私保护的同时发布它们一直是一项具有挑战性的任务。在本文中，我们提出了一种在隐私保护轨迹发布中保护敏感地点访问的算法。通过使用敏感区域概括敏感地点，并基于差分隐私扭曲敏感区域内的子轨迹，我们的方法不仅可以防止敏感地点访问的泄漏，还可以保留个人运动信息。它包含两个关键组件。首先，我们根据人类移动模式和移动模型在敏感地点周围生成敏感区域。敏感区域的形成方式使得对手的背景知识不会增加对手对轨迹是否已在敏感位置停止的信念。其次，为了防止敏感区域内过多的个人运动信息丢失和敏感地点访问泄漏，我们从其中的子轨迹中选择可靠段，将可靠段建模为探索树，并合成ɛ-差异私有子轨迹. 我们在真实世界数据集上的实验表明，我们的方法提供了良好的实用性，我们的子轨迹合成方法保留了个人运动的详细信息。为了防止敏感区域内过多的个人运动信息丢失和敏感地点访问泄漏，我们从其中的子轨迹中选择可靠段，将可靠段建模为探索树，并合成ɛ-差异私有子轨迹。我们在真实世界数据集上的实验表明，我们的方法提供了良好的实用性，我们的子轨迹合成方法保留了个人运动的详细信息。为了防止敏感区域内过多的个人运动信息丢失和敏感地点访问泄漏，我们从其中的子轨迹中选择可靠段，将可靠段建模为探索树，并合成ɛ-差异私有子轨迹。我们在真实世界数据集上的实验表明，我们的方法提供了良好的实用性，我们的子轨迹合成方法保留了个人运动的详细信息。