Privacy-preserving mechanism for trajectory data publishing based on deep generative models
-
摘要: 为克服当前轨迹数据发布中轨迹数据质量欠佳和隐私保护不足等问题，提出了一种基于深度生成模型的轨迹数据发布隐私保护机制；通过结合时间、距离和速度等多维度特征提取轨迹停留点，对车辆的原始轨迹进行分段，从而降低数据冗余与模型训练复杂度；为有效捕捉轨迹数据中的时空特征，运用长短期记忆网络并结合自注意力机制，设计了一种基于生成对抗网络的轨迹合成模型；利用长短期记忆网络和自注意力机制对轨迹序列进行学习，再结合生成对抗网络模型进行训练以生成高质量的合成轨迹；为进一步增强轨迹的个性化隐私保护，应用双向门控循环单元设计了面向用户的轨迹预测模型，并对用户历史轨迹信息进行训练，通过学习-预测的模式，从训练数据中挖掘分析用户的出行规律，形成个性化的用户轨迹预测模型；通过轨迹预测模型对合成轨迹进行分段预测，根据预测结果，识别需要进一步进行强化隐私保护的轨迹段，并添加差分隐私噪声，提升隐私保护，从而获得用于数据发布的隐私保护轨迹。仿真试验结果表明：与现有方法相比，在西安出租车和重卡轨迹数据场景下，均方根误差值降低至26 m，JS散度值在空间分布和时间分布上分别降低至0.12和0.19，互信息值降低至1.97。提出的轨迹数据保护机制在轨迹可用性和隐私保护性能方面均有显著提升，证明了该机制在隐私保护和数据效用之间的良好平衡。Abstract: In order to overcome the problems such as poor trajectory data quality and insufficient privacy preservation in the trajectory data publishing, a privacy-preserving mechanism for trajectory data publishing based on deep generative models was proposed. Trajectory stop points were extracted by integrating multi-dimensional features such as time, distance, and speed, and the raw vehicle trajectories were segmented to reduce data redundancy and model training complexity. To effectively capture the spatio-temporal features in trajectory data, a trajectory synthesis model based on a generative adversarial network was designed by applying a long short-term memory network combined with a self-attention mechanism. The trajectory sequences were learned using a long short-term memory network and a self-attention mechanism, and then the model was trained with a generative adversarial network to generate high-quality synthetic trajectories. To further enhance the personalized privacy preservation of trajectories, a trajectory prediction model for users was designed by applying a bidirectional gated recurrent unit, and the model was trained with users' historical trajectory information. Through the learning and prediction mode, users' travel patterns were explored and analyzed from the training data to form personalized user trajectory prediction models. The synthetic trajectories were segmented and predicted by the trajectory prediction model. According to the prediction results, the trajectory segments requiring further enhanced privacy preservation were identified, with differential privacy noise added to improve privacy preservation, so as to obtain privacy-preserving trajectories for data publishing. Simulation results show that compared with existing methods, in the scenarios of taxi in Xi'an city and heavy truck trajectory data, the root-mean-square error reduces to 26 m. The JS divergences in spatial and temporal distributions reduce to 0.12 and 0.19, respectively, and the mutual information score reduces to 1.97. The proposed trajectory data preservation mechanism has been significantly improved in terms of trajectory availability and privacy preservation performance, demonstrating a good balance between privacy preservation and data utility.
-
图 5 时间阈值和长停留点数量相关性
Figure 5. Correlation between time threshold and number of long dwell points
图 6 距离阈值和徘徊点数量相关性
Figure 6. Correlation between distance threshold and number of wandering points
图 7 速度阈值和徘徊点数量相关性
Figure 7. Correlation between speed threshold and number of wandering points
图 9 不同预测距离阈值下数据可用性评估
Figure 9. Evaluation of data availability under different prediction distance thresholds
图 10 不同预测距离阈值下隐私保护性评估
Figure 10. Evaluation of privacy protection under different prediction distance thresholds
图 11 不同预测距离阈值下综合评估
Figure 11. Comprehensive evaluation under different prediction distance thresholds
图 17 重卡数据集可用性对比
Figure 17. Comparative analysis of dataset availability for heavy-duty trucks
图 18 停留点密集数据集均方根误差对比
Figure 18. Comparison of root mean square errors for stay-point dense datasets
图 20 重卡数据集可用性对比
Figure 20. Comparative analysis of availability of heavy-duty truck dataset
图 22 出租车数据集隐私保护程度对比
Figure 22. Comparative analysis of privacy protection levels in taxi datasets
图 23 重卡数据集隐私保护程度对比
Figure 23. Comparative analysis of privacy protection levels in heavy-duty truck datasets
图 24 停留点密集数据集互信息对比
Figure 24. Comparison of mutual informations for stay-point dense datasets
表 1 试验参数
Table 1. Experimental parameters
参数 取值 距离阈值/m 100 时间阈值/s 300 速度阈值/(m·s-1) 1 LSTM单元数量 100 α 0.3 β 0.5 Ψ 0.2 
下载: 导出CSV
-
[1] WANG W, WANG Y, DUAN P, et al. A triple real-time trajectory privacy protection mechanism based on edge computing and blockchain in mobile crowdsourcing[J]. IEEE Transactions on Mobile Computing, 2022, 22(10): 5625-5642. [2] ZHAO Y, CHEN J. Vector-indistinguishability: location dependency based privacy protection for successive location data[J]. IEEE Transactions on Computers, 2023, 73(4): 970-979. [3] XING L, LI B, LIU L, et al. Trajectory privacy protection method based on sensitive semantic location replacement[J]. Computer Networks, 2024, 250: 1-10. [4] JIN F, HUA W, FRANCIA M, et al. A survey and experimental study on privacy-preserving trajectory data publishing[J]. IEEE Transactions on Knowledge and Data Engineering, 2022, 35(6): 5577-5596. [5] XI J, SHI M, ZHANG W, et al. Trajectory privacy-protection mechanism based on multidimensional spatial-temporal prediction[J]. Symmetry, 2024, 16(9): 1-16. [6] HU P, CHU X, ZUO K, et al. Security-enhanced data sharing scheme with location privacy preservation for internet of vehicles[J]. IEEE Transactions on Vehicular Technology, 2024, 73(9): 13751-13764. doi: 10.1109/TVT.2024.3393302 [7] 蒋伟进, 王海娟, 周为, 等. 基于自适应连续时间的群智感知轨迹隐私保护方案[J]. 电子学报, 2023, 51(10): 2894-2901.JIANG Wei-jin, WANG Hai-juan, ZHOU Wei, et al. Track privacy protection scheme based on adaptive continuous time in crowdsensing[J]. Chinese Journal of Electronics, 2023, 51(10): 2894-2901. [8] SCHESTAKOV S, GOTTSCHALK S, FUNKE T, et al. RE-Trace: Re-identification of modified GPS trajectories[J]. ACM Transactions on Spatial Algorithms and Systems, 2024, 10(4): 1-28. [9] GRAMAGLIA M, FIORE M, FURNO A, et al. GLOVE: Towards privacy-preserving publishing of record-level-truthful mobile phone trajectories[J]. ACM/IMS Transactions on Data Science (TDS), 2021, 2(3): 1-36. [10] SHEN H, WANG Y, ZHANG M. A privacy-preserving trajectory publishing method based on multi-dimensional sub-trajectory similarities[J]. Sensors, 2023, 23(24): 1-22. doi: 10.1109/JSEN.2023.3338435 [11] WU W, SHANG W, LEI R, et al. A trajectory privacy protect method based on location pair reorganization[J]. Wireless Communications and Mobile Computing, 2022, 2022(1): 1-16. [12] QIU S, PI D, WANG Y, et al. Novel trajectory privacy protection method against prediction attacks[J]. Expert Systems with Applications, 2023, 213: 1-14. [13] GAO Z, HUANG Y, ZHENG L, et al. Protecting location privacy of users based on trajectory obfuscation in mobile crowdsensing[J]. IEEE Transactions on Industrial Informatics, 2022, 18(9): 6290-6299. [14] WANG H, ZHANG Z, WANG T, et al. PrivTrace: differentially private trajectory synthesis by adaptive markov models[C]//USENIX Association. 32nd USENIX Security Symposium (USENIX Security 23). Anaheim: USENIX, 2023: 1649-1666. [15] CHENG W, WEN R, HUANG H, et al. OPTDP: towards optimal personalized trajectory differential privacy for trajectory data publishing[J]. Neurocomputing, 2022, 472: 201-211. [16] GU Z, ZHANG G. Trajectory data publication based on differential privacy[J]. International Journal of Information Security and Privacy, 2023, 17(1): 1-15. [17] XU C, ZHU L, LIU Y, et al. DP-LTOD: differential privacy latent trajectory community discovering services over location-based social networks[J]. IEEE Transactions on Services Computing, 2018, 14(4): 1068-1083. [18] WU L, QIN C, XU Z, et al. TCPP: achieving privacy-preserving trajectory correlation with differential privacy[J]. IEEE Transactions on Information Forensics and Security, 2023, 18: 4006-4020. [19] 陈思, 付安民, 苏铓, 等. 基于差分隐私的轨迹隐私保护方案[J]. 通信学报, 2021, 42(9): 54-64.CHEN Si, FU An-min, SU Mang, et al. Trajectory privacy protection scheme based on differential privacy[J]. Journal of Communications, 2021, 42(9): 54-64. [20] 朱素霞, 刘抒伦, 孙广路. 基于相对熵和K-means的形状相似差分隐私轨迹保护机制[J]. 通信学报, 2021, 42(2): 113-123.ZHU Su-xia, LIU Shu-lun, SUN Guang-lu, et al. Shape similarity differential privacy trajectory protection mechanism based on relative entropy and K-means[J]. Journal of Communications, 2021, 42(2): 113-123. [21] 姚俊峰, 何瑞, 史童童, 等. 基于机器学习的交通流预测方法综述[J]. 交通运输工程学报, 2023, 23(3): 44-67. doi: 10.19818/j.cnki.1671-1637.2023.03.003YAO Jun-feng, HE Rui, SHI Tong-tong, et al. Review on machine learning-baed traffic flow prediction methods[J]. Journal of Traffic and Transportation Engineering, 2023, 23(3): 44-67. doi: 10.19818/j.cnki.1671-1637.2023.03.003 [22] 叶阿勇, 孟玲玉, 赵子文, 等. 基于预测和滑动窗口的轨迹差分隐私保护机制[J]. 通信学报, 2020, 41(4): 123-133.YE A-yong, MENG Ling-yu, ZHAO Zi-wen, et al. Trajectory differential privacy protection mechanism based on prediction and sliding window[J]. Journal of Communications, 2020, 41(4): 123-133. [23] 由林麟, 贺俊姝, 陈坤旭, 等. 面向个体出行推荐的联邦异质性模型和算法[J]. 交通运输工程学报, 2023, 23(5): 253-263. doi: 10.19818/j.cnki.1671-1637.2023.05.018YOU Lin-lin, HE Jun-shu, CHEN Kun-xu, et al. Federated heterogeneous model and algorithm for personal travel recommendation[J]. Journal of Traffic and Transportation Engineering, 2023, 23(5): 253-263. doi: 10.19818/j.cnki.1671-1637.2023.05.018 [24] 崔建勋, 要甲, 赵泊媛. 基于深度学习的短期交通流预测方法综述[J]. 交通运输工程学报, 2024, 24(2): 50-64. doi: 10.19818/j.cnki.1671-1637.2024.02.003CUI Jian-xun, YAO Jia, ZHAO Po-yuan. Review on short-term traffic flow prediction methods based on deep learning[J]. Journal of Traffic and Transportation Engineering, 2024, 24(2): 50-64. doi: 10.19818/j.cnki.1671-1637.2024.02.003 [25] CHOI S, KIM J, YEO H. TrajGAIL: generating urban vehicle trajectories using generative adversarial imitation learning[J]. Transportation Research Part C: Emerging Technologies, 2021, 128: 1-25. [26] CHEN X, XU J, ZHOU R, et al. TrajVAE: a variational AutoEncoder model for trajectory generation[J]. Neurocomputing, 2021, 428: 332-339. [27] HU J, HE J, ZHU N, et al. Trajectory privacy preservation model based on LSTM-DCGAN[J]. Future Generation Computer Systems, 2025, 163: 1-13. [28] JIANG Y, WU Y, ZHANG S, et al. Fedvae: trajectory privacy preserving based on federated variational autoencoder[C]//IEEE. 2023 IEEE 98th Vehicular Technology Conference (VTC2023-Fall). New York: IEEE, 2023: 1-7. [29] WANG X, LIU X, LU Z, et al. Large scale GPS trajectory generation using map based on two stage GAN[J]. Journal of Data Science, 2021, 19(1): 126-141. [30] CAO C, LI M. Generating mobility trajectories with retained data utility[C]//ACM. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. New York: ACM, 2021: 2610-2620. [31] SHIN J, SONG Y, AHN J, et al. TCAC-GAN: synthetic trajectory generation model using auxiliary classifier generative adversarial networks for improved protection of trajectory data[C]//IEEE. 2023 IEEE International Conference on Big Data and Smart Computing (BigComp). New York: IEEE, 2023: 314-315. [32] CAO X, YU J, HAN J, et al. A transformer decoder-based generative adversarial model with trajloss function for privacy-preserving trajectory publishing[C]//ACM. Proceedings of the 2022 5th International Conference on Machine Learning and Natural Language Processing. New York: ACM, 2022: 271-278. [33] ZHANG J, HUANG Q, HUANG Y, et al. DP-TrajGAN: a privacy-aware trajectory generation model with differential privacy[J]. Future Generation Computer Systems, 2023, 142: 25-40. [34] SUN X, WO T. A privacy-preserving and research-utilizable trajectory generator via deep generative approach[C]//IEEE. 2023 6th International Conference on Electronics Technology (ICET). New York: IEEE, 2023: 1017-1021. [35] KIM J W, JANG B. Deep learning-based privacy-preserving framework for synthetic trajectory generation[J]. Journal of Network and Computer Applications, 2022, 206: 1-10. -
点击查看大图
计量
- 文章访问数: 176
- HTML全文浏览量: 64
- PDF下载量: 14
- 被引次数: 0
