Trajectory Protection Mechanism Based On Differential Privacy

In recent years,people pay more and more attention to personal privacy,and trajectory data,as a kind of data containing rich spatio-temporal information of users,has also attracted people's attention.Trajectory protection mechanism based on differential privacy has gradually become a research hotspot,but the privacy budget with fixed differential privacy will lead to the waste of privacy budget;and the smaller the privacy budget is set,the worse the data availability of trajectory protection mechanism exists.In order to reduce the waste of privacy budget and improve the usability of trajectory,this paper further optimizes the location privacy level algorithm from the point of interest and trajectory shape to reduce the waste of privacy budget and improve the usability of trajectory.DPPOI(Differential privacy Based on DBSCAN and point of interest)is proposed in this paper.The mechanism of the main idea is to a user's all trajectory data form history database,DBSCAN algorithm is used to cluster the historical trajectory data set clustering,extract the set point of interest,calculating the distance between the points and interest measure trajectory to a certain point of interest orientation variance,and the basis of the intensity of the bias level of privacy,set the path points of the last points to the location of different level of privacy to take different intensity of disturbance.According to the experimental results,the DPPOI mechanism has good trajectory availability in the historical trajectory data set with sparse data points.In order to further improve the trajectory usability,this paper proposes the DPKTS(Differential Privacy Based on K-means similarity)trajectory protection mechanism based on relative entropy and K-means.The DPKTS mechanism uses relative entropy to measure the privacy level of location points.The privacy level allocation scheme sets the ideal state,and takes the information entropy of the ideal state as the standard to calculate the relative entropy of the location point set and the ideal state as the privacy level.Due to setting the ideal state,the privacy level algorithm will try to approach the ideal state when allocating the privacy budget to all the historical trajectory data sets,so it has a good performance for all the historical trajectory data sets.In addition,DPKTS mechanism uses trajectory shape similarity to improve trajectory availability.By using Fréchet distance and K-means algorithm,the release location most similar to the real trajectory can be effectively selected to improve trajectory availability.The experimental results show that both DPPOI and DPKTS mechanisms improve track availability,and DPKTS mechanisms have better performance in track availability.