Location Privacy Protection Based On Geo-indistinguishability For Crowd Sensing

The rapid iteration of Internet applications and intelligent terminals provides a software and hardware foundation and a huge user base for Mobile Crowdsensing.Location-Based Service is one of the most closely used services in Crowdsensing,such as commentary,taxis,and social matching.However,data leakage incidents often occur,and users are particularly concerned about their location privacy due to their distrust of servers.Location privacy protection is of great significance,and how to reduce the risk of user location privacy leakage while ensuring Qo S of Crowdsensing is focus of recent researches.Therefore,it is necessary to provide suitable location privacy protection strategies for various users and different scenarios.This thesis analyzes and compares different privacy-preserving strategies,and proposes a location privacy protection mechanism based on Geo-indistinguishability for Crowdsensing scenarios,which is a localized differential privacy framework.It is indistinguishable from the real location and the obfuscated location within the scope of action for servers.Based on this framework,three specific strategies are designed in different scenarios:1)Aiming at the LBS service requested by independent location,the user’s high-frequency access area exposes more side privacy,this research proposes the Privacy Budget Redistribution Algorithm based on Heat Maps.Specifically,this thesis exploits the density-based clustering method to analyze the user’s historical check-in records,divide the user’s hotspot area and redistribute the privacy budget under the premise of ensuring the same average privacy budget to achieve noise from hotspots to non-hotspots.Diffusion to increase the level of privacy protection in hotspots.Experiments on the data set show the effectiveness of the algorithm on privacy protection.2)Targeting at the scenario where the prior probability is leaked and the server has the ability to reason and attack after mastering the perturbation mechanism,this thesis proposes an adaptive noise obfuscation strategy based on the Stackelberg game.The user acts as the leader,the server uses the Bayesian attack as the follower,and both parties dynamically choose the perturbation and attack strategies.Under the dual effects of geographic indistinguishability and inference error,a linear scale model is established with the goal of minimum the utility loss,and a location obfuscation strategy is adaptively selected.The simulation results show that on the premise of ensuring the privacy effect,this strategy can greatly reduce the loss of utility and improve the Qo S of Crowdsensing.3)For time-series-related trajectory data release scenarios,adding uniform and independent noise to locations will enormously reduce data availability.This thesis proposes a personalized trajectory protection strategy for time-series-related noise disturbances.Semantic segmentation is performed on the map,the user sets sensitive areas individually,analyzes historical trajectories,conducts personalized privacy evaluation for the location points in the to-be-released trajectories based on sensitivity and activity,and determines the level of privacy protection based on the scores.Then,through a special set filter,the random noise sequence generated by the personalized evaluation is combined into a time-series-related noise sequence and the original track to form a release track,which protects the time-series correlation of the track data.Experiments show that this strategy improves the privacy-preserving performance and data query quality.