Scholarlink

<p>To address the challenge of privacy and usability in trajectory data, this paper proposes a Hasse Diagram Sensitivity Differential Privacy and Reinforcement Learning Mechanism (HSLRL-DP). Firstly, for the temporal information in trajectories, a Density Peak Clustering Algorithm with Consistency Distance Measure (DPCCD) is employed to identify the stopping points within the trajectories. Secondly, a Hasse Diagram Sensitive Position Calculation Algorithm (HDSC) is introduced to store the clustered trajectories, constructing a partial order relation based on the TF-IDF values of the stopping points to calculate sensitive positions. Lastly, using the Boundary Laplace Mechanism (BLM), the optimal Laplace boundary is calculated through reinforcement learning, and bounded Laplace noise is added to the Hasse Diagram, storing sensitive positions to achieve differential privacy. Comparative experiments were conducted using the Geolife and T-Drive datasets, and the results demonstrate that the HSLRL-DP mechanism proposed in this paper offers superior performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

HSLRL-DP: based on Hasse sensitivity levels and reinforcement learning trajectory privacy enhancement mechanism

  • 期刊论文

  • 发表日期：

    2025年8月19日

查看全文
  • Yanmei Shen,
  • Yabiao Sang,
  • Hui Wang,
  • Zihao Shen,
  • Peiqian Liu

摘要

To address the challenge of privacy and usability in trajectory data, this paper proposes a Hasse Diagram Sensitivity Differential Privacy and Reinforcement Learning Mechanism (HSLRL-DP). Firstly, for the temporal information in trajectories, a Density Peak Clustering Algorithm with Consistency Distance Measure (DPCCD) is employed to identify the stopping points within the trajectories. Secondly, a Hasse Diagram Sensitive Position Calculation Algorithm (HDSC) is introduced to store the clustered trajectories, constructing a partial order relation based on the TF-IDF values of the stopping points to calculate sensitive positions. Lastly, using the Boundary Laplace Mechanism (BLM), the optimal Laplace boundary is calculated through reinforcement learning, and bounded Laplace noise is added to the Hasse Diagram, storing sensitive positions to achieve differential privacy. Comparative experiments were conducted using the Geolife and T-Drive datasets, and the results demonstrate that the HSLRL-DP mechanism proposed in this paper offers superior performance.