基于BIGRU的轨迹数据发布隐私保护方案

Privacy protection scheme for trajectory data publishing based on BIGRU

在保证轨迹数据发布隐私性的同时,需要提高发布数据的可用性,将机器学习算法应用于轨迹数据处理可以提高轨迹数据的可用性。针对轨迹数据发布可用性问题,提出一种面向轨迹数据发布的结合双向门控循环单元(bidirectional gated recurrent unit,BIGRU)和差分隐私(differential privacy,DP)的轨迹隐私保护方案。通过应用BIGRU对轨迹数据进行预处理从而提高轨迹数据的可用性,对轨迹数据进行聚类泛化并使用差分隐私指数机制进行分区选择从而达到了隐私保护的目的,将得到的泛化轨迹数据集进行异常处理并发布。仿真实验结果表明,该方案不仅具有较好的数据可用性,也有一定的效率优势。

While ensuring the privacy of trajectory data publication,it is necessary to improve the usability of the published data.Applying machine learning algorithms to trajectory data processing can improve the usability of trajectory data.A trajectory privacy protection scheme is proposed to solve the usability problem in trajectory data publishing,which combines bidirectional gated recurrent unit(BIGRU)and differential privacy(DP)for trajectory data publishing.Firstly,by applying BIGRU to preprocess trajectory data,the usability of trajectory data is improved.Then,the trajectory data is clustered and generalized,and a differential privacy index mechanism is used for partition selection to achieve privacy protection.Finally,the obtained generalized trajectory data set is subjected to exception handling and then gets published.Simulation results show that this scheme not only has good data usability,but also has certain efficiency advantages.