【摘要】 针对现有轨迹数据隐私保护方案在处理轨迹数据中的稀疏区域时，对数据可用性影响较大的问题。提出了一种基于密度感知和差分隐私的轨迹隐私保护方案。将轨迹数据集按照不同时间戳进行初始聚类，并对聚类质心应用拉普拉斯机制注入噪声，有效防止数据隐私泄露。在此基础上根据聚类密度阈值，提出一种密度感知分级聚类方法，以精细化处理稀疏区域，提升轨迹数据的可用性。综合选取加噪后的初始聚类质心和分级聚类质心作为轨迹代表点，生成隐私保护的泛化轨迹，以有效抵御轨迹重构攻击和推断攻击。实验结果表明，在稀疏轨迹数据场景下，该方案在保证轨迹数据隐私性的同时显著提高数据可用性。更多还原

【Abstract】 To address the issue of reduced data utility in sparse regions caused by existing trajectory data privacy protection schemes, a trajectory privacy protection scheme based on density awareness and differential privacy was proposed. The trajectory data set was initially clustered according to different timestamps, and the Laplace mechanism was applied to inject noise into the cluster centroids, effectively preventing data privacy leakage. Based on a clustering density threshold, a density-aware hierarchical clustering method was then introduced to refine the processing of sparse regions and enhance the utility of trajectory data. A combination of the noise-injected initial cluster centroids and the hierarchical clustering centroids was selected as trajectory representative points to generate privacy-preserving generalized trajectories, effectively resisting trajectory reconstruction and inference attacks. Experimental results demonstrate that, in sparse trajectory data scenarios, the proposed scheme ensures trajectory data privacy while significantly improving data utility.更多还原