We describe a new generative algorithm called trajectory generative mechanism (TGM) for publishing trajectory datasets with $\varepsilon $ -differential privacy guarantee, which achieves substantially higher computational efficiency and utility (practical) than the state-of-the-art algorithms. Our algorithm first encodes (models) the data as a graphical generative model and accurately captures the statistics of moving object trajectories. Using this model, TGM then privately generates synthetic trajectories such that the noise is optimally added to capture the movement direction of an object. Our algorithm preserves both the spatial and temporal information of trajectories in the generated dataset, requires less memory and computation than the competing approaches, and preserves the properties of real trajectory data in terms of traveled distance and stay location. We demonstrate the performance of TGM on both real and simulated datasets with a wide range of settings. Our experimental results show that TGM achieves high utility and efficiency by using the properties of the data.

中文翻译：

---

TGM：一种发布具有差异隐私的轨迹的生成机制


我们描述了一种称为轨迹生成机制（TGM）的新生成算法，用于发布具有 $\varepsilon $ 差分隐私保证的轨迹数据集，该算法比最先进的算法具有更高的计算效率和实用性（实用）。我们的算法首先将数据编码（建模）为图形生成模型，并准确捕获移动物体轨迹的统计数据。然后，TGM 使用该模型私下生成合成轨迹，以便以最佳方式添加噪声以捕获对象的移动方向。我们的算法保留了生成的数据集中轨迹的空间和时间信息，比竞争方法需要更少的内存和计算，并保留了真实轨迹数据在行驶距离和停留位置方面的属性。我们在各种设置的真实数据集和模拟数据集上展示了 TGM 的性能。我们的实验结果表明，TGM 通过利用数据的特性实现了高实用性和高效率。