基于自适应连续时间的群智感知轨迹隐私保护方案

Track Privacy Protection Scheme Based on Adaptive Continuous Time in Crowdsensing

针对轨迹差分隐私保护存在的预测精度差、隐私预算分配效用低的问题,本文提出自适应连续时间序列下的群智感知轨迹预测方案.首先在任务分配阶段,为参与者分配轨迹路线;其次引入隐马尔可夫模型(Hidden Markov Model,HMM),对轨迹进行预测;然后使用预分配和自适应分配相结合的综合隐私预算分配方法,降低隐私预算;最后利用拉普拉斯机制,进行位置扰动.实验结果表明,与相关工作相比,所提方法兼顾预测性和低预算性,对群智感知中参与者在轨迹隐私安全保护上具有良好的保护效果.

In response to the problems of poor prediction accuracy and low utility of privacy budget allocation in tra⁃jectory differential privacy protection,our paper proposes an adaptive trajectory prediction scheme for continuous time se⁃ries in crowdsensing.Firstly,in the task assignment phase,trajectory routes are assigned to participants.Then,the HMM(Hidden Markov Model)is introduced to predict the trajectories.Next,a comprehensive privacy budget allocation method combining pre-allocation and adaptive allocation is used to reduce the privacy budget.Finally,the laplace mechanism is ap⁃plied to perturb the locations.Experimental results show that compared with related work,the proposed method achieves a balance between prediction accuracy and low budget requirements,and provides good privacy protection for participants in trajectory privacy security in crowdsensing.