| With the proliferation of smartphones,tablets and other Internet of Things(IoT)devices,location-based services(LBS)have become increasingly popular and are shaping the way we use the internet,which play a hugely important role in social life.This has also made it more common for users’ location data to be collected through the use of mobile phones,GPS devices and geo-tagged social media,with huge amounts of user location data being stored on cloud-based servers.This vast amount of location data is of great academic and commercial value,but it also poses the risk of breaches of privacy for users.Existing research has mainly focused on the spatial dimension of location data,with little consideration of other information.Some research solutions only work in specific contexts and do not have generality.Some are one-size-fits-all approaches to protecting location privacy of users which do not take into account the differences in the sensitivity of different location data and the privacy needs of different users,making it difficult to adapt to complex and changing scenarios.In response to these shortcomings,Privacy technologies in location based services is investigated,and the work and innovations are shown below.Firstly,to address the problem that most current real-time LPPM(location privacy protection mechanism)do not adequately consider the background knowledge possessed by attackers,a multidata-based location privacy protection scheme MDLS is proposed by considering the query probability,semantic information and physical distribution of locations.The locations whose query probability is close to that of the user’s real location are selected based on the max heap to form a dummy location candidate set.The physical distance and semantic distance between the candidate locations and the real location are calculated to filter out the physically dispersed and semantically diverse fake locations.A dummy location set of size k containing the real location of the user is generated.The experimental results show that the proposed algorithm can quickly and effectively generate a physically dispersed and semantically diverse set of dummy locations.Secondly,a multi-dimensional feature fake trajectory generation model MFF-TrajGAN is given,which combines the fake location technique with a generative adversarial network model,to generate privacy-preserving synthetic trajectory data.Firstly,a trajectory encoding model is designed based on the multi-dimensional features of the trajectory data to convert the original trajectory into the specific format encoding.Secondly,the trajectory generator acquires the data distribution and patterns of historical trajectory data and generates synthetic trajectory data that cannot be recognised by the trajectory discriminator based on their corresponding initial trajectory data and random noise.The experimental results show that our model can better prevent users from being re-identified than other common traditional privacy-preserving methods,and also retains the basic spatial,temporal and semantic features of the actual trajectory data,which can effectively balance the privacy and practicality of the published trajectory data.Finally,a location privacy-preserving location publishing system for users and LBS servers is proposed,which combines the multivariate-data fake location screening algorithm MDLS with the multidimensional feature fake trajectory generation algorithm MFF-TrajGAN. |
