【摘要】 位置信息作为数字时代重要的个人数据资产，在提供便利服务的同时面临着严重的隐私泄露风险。本地化差分隐私模型无须依赖可信的第三方，近年来受到广泛关注。然而，现有本地化差分隐私位置保护方法普遍面临两大挑战：一是空间分割结构难以适应位置分布的复杂特性；二是用户与服务器端的通信与计算开销制约了系统效率。针对上述挑战，提出一种基于关键路径的动态本地化差分隐私位置保护方法。该方法首先通过非均匀四叉树空间分割与Hilbert曲线遍历，构建适配于用户分布密度的空间索引，有效提升了数据可用性。继而由服务器运行提出的关键路径编码机制，将复杂的分割结构压缩为简洁的路径信息，减少了下发参数时的通信开销。用户端根据本地化差分隐私模型对自身所处区域的Hilbert索引编码进行随机响应扰动，实现原始位置的隐私保护。服务器端则根据收集到的用户扰动位置编码进行聚合分析，并依据位置分布的时空连续性执行所提出的空间分割结构动态调整策略，以极低的计算成本高效适应用户分布的动态变化。在实际位置数据集上的实验证明，所提方法能够在实现用户位置本地化差分隐私保护的基础上，提高数据可用性和算法运行效率。更多还原

【Abstract】 Location information was recognized as a critical personal data asset in the digital age, offering convenient services while simultaneously posing significant risks of privacy breaches. Local differential privacy models, which do not rely on trusted third parties, had garnered widespread attention. However, significant challenges were identified in existing location protection methods, including the difficulty of adapting spatial partitioning to complex location distributions, along with high communication and computational overhead that limited system efficiency. To address these challenges, a dynamic local differential privacy location protection method based on a critical path was proposed. A spatial index adapted to user distribution density was constructed through non-uniform quadtree spatial partitioning and Hilbert curve traversal, which effectively improved data usability. Subsequently, the proposed critical path encoding mechanism was executed on the server side to compress the complex partition structure into concise path information, thereby reducing communication overhead during parameter transmission. On the user side, the Hilbert index encoding of the user’s region was perturbed using a randomized response mechanism under the local differential privacy model to protect the privacy of the original location. On the server side, the collected perturbed location encodings from users were aggregated and analyzed. Based on the spatiotemporal continuity of location distribution, the proposed spatial partition structure dynamic adjustment strategy was then implemented to efficiently adapt to dynamic changes in user distribution at an extremely low computational cost. Experiments conducted on real-world location datasets demonstrate that the proposed method provides improved location data availability and algorithm runtime efficiency while achieving local differential privacy protection for user locations.更多还原