Research On Privacy Preservation In The Publication Of Trajectories

Recent years, with the widespread use of GPS and other mobile-based devices and services, the number of trajectories collected by service providers is continuously increasing.On one hand, trajectories contain a lot of valuable information; it can help to support a mobile-related strategic decision by mining and analyzing these spatiotemporal trajectories.On the other hand, the adversary can use knowledge of each trajectory to identify users'sensitive information, such as hobbies, behavioral patterns, customs and health conditions,and lead to an unpredictable harm. Therefore, before publishing trajectories, it is necessary to process original data to prevent the user's privacy to be identified.At present, many researchers have launched studies to solve this problem and have made some achievements. Trajectory k-anonymity which is more commonly used is a relatively good way to solve this problem. Traditional trajectory k-anonymity method is to anonymize at least k trajectories that are nearest on both the time and space, in order to make these k trajectories indistinguishable. The anonymity is about all points on the trajectory, this leads to a serious information distortion of trajectory, and affecting data quality. Meanwhile, it's necessary to cluster trajectories in order to search for k-anonymous set. However, since the distribution of trajectories is difficult to unify both in time and space, more sophisticated algorithms are required to preprocessing.In addition, the privacy requirements of each point on the trajectory are different, and the background knowledge of adversaries is relevant to some special points. In allusion to above problems, this paper puts forward to the points of interest-based trajectory k-anonymity privacy protection method in publication of trajectories. This method is mainly by protecting points of interest (POIs) to achieve protecting trajectory privacy. Apparently, it needs to extract POIs first before trajectory anonymity. Firstly, we give the definition of POI based on the privacy requirements and research needs. Secondly, we avoid the complex trajectory distance calculation on the process of searching k-anonymous set. Instead, we divide similar trajectories into one group according to their features, that is similar trajectories will pass the same region of interest (ROI). Finally, trajectories in each k-anonymous set are publication.Here the idea of swapping locations is adopted. The anonymity of trajectories is only about points of interest, and we will do nothing for each ordinary point. Therefore, in this paper,trajectory k-anonymity is about the trajectories which are reconstructed by points of interest.We do experiments on the real world data set, and evaluate the quality of published data by two different ways. By comparing to existing methods, it is proved that our method can improve the data quality on the condition of achieving privacy requirements.