Research On User Trajectory Privacy Protection Based On Differential Privacy

In recent years,with the development of smart mobile devices and network communications,mobile crowdsensing(MCS)has become a popular data collection method.However,the contradiction between the privacy issues of participants and the accuracy of the collected data has restricted the further development of MCS.The thesis is mainly focused on the security of participants' trajectories in MCS.At present,there are lots of works on participants' trajectories privacy protection,and most of these works are based on anonymity to protect user's real trajectory information from being leaked.However,this method is only based on semantic,which unable to provide a strict privacy definition,meanwhile,it is sensitive to the background knowledge possessed by the adversary.In addition,some researchers have applied differential privacy(DP)to guarantee trajectory privacy,but the implementation of differential privacy requires noise,which still has conflicts between data accuracy and participants' privacy.Due to the shortcoming in current works,this thesis proposes a new type of trajectory privacy protection scheme based on differential privacy,which can guarantee the participants' trajectory privacy while ensuring the accuracy of data.The main works are as follows:(a)We carried out detailed investigations on related works especially on anonymous and differential privacy based location privacy protection,and then we compared these works' innovations and shortcomings,which would be good reference for our work.(b)We propose a trajectory privacy protection scheme based on differential privacy.Markov model is used to quantify the probability of participants' trajectories,and adopt differential privacy to filter road segments that have little effects on the posterior probability of the real trajectory for reporting against Bayesian attack model.At the same time,some metrics are defined to measure the level of privacy protection.Experiments show that our scheme can perform well.(c)The work of the second part is further optimized,and similar trajectories are used to obfuscate the real path,so as to reduce the risk of real trajectory being identified,thereby achieving differential privacy.More over,the segments in the map are divided into sensitive segments and non-sensitive segments,thus reducing the leakage of more private information in the reported sections.