Differential Privacy Protection Of Worker Locations In Spatial Crowdsourcing

With the rapid development of mobile Internet technology,smart phones have gradually surpassed PC as the most common network access device in the world,playing an increasingly important role in human life.Mobile devices provide a variety of services for people,but also collect a variety of privacy information.Illegal elements may make targeted privacy attacks through data analysis and processing,which will affect the life and property security of mobile device users.In particular,user's location privacy has attracted more and more attention from academic and commercial institutions,and related studies have great academic value and practical significance.Spatial Crowdsourcing(SC)is a crowdsourcing form based on worker's location information.Workers need to go to designated locations to perform tasks.Typical SC tasks include taking photos,recording videos,reporting temperature,etc.As long as workers have mobile devices,they can accept SC tasks,which will greatly improve the efficiency of their work.Workers need to report their location information to the server if they want to undertake the task.However,the server is usually an untrustworthy entity.Once the location information of SC workers is illegally collected and shared,it may expose sensitive information such as personal health,political and religious preferences.For this,one can add some noises by using differential privacy technology to protect individual private information in datasets.Related studies have attracted much attention in recent years.This thesis focuses on the differential privacy protection of SC worker's location privacy,and in particular,this thesis proposes a solution for SC worker's location privacy protection.Our main work and contributions include the following two parts:(1)This thesis proposes a spatial crowdsourcing worker location privacy protection scheme based on adaptive region partitioning.This thesis considers how to ensure the success rate of task acceptance while protecting the location privacy of workers.This thesis designs a novel adaptive workspace partitioning method,and analyzes and optimizes the partitioning granularity selection theoretically and experimentally.This thesis first proposes a partitioning method with historical data learning.In terms of historical data with added noises,a technique of regression analysis in machine learning is used to make predictions at two levels and perform adaptive grid partitioning.In the task release area construction stage,taking into account the divided work space,the worker density of each grid,the distance between the task and the grid,etc.,assign a score to each grid,and use the exponential mechanism to construct the task release area,so that the score Higher areas have a higher probability of being selected,which is also in line with our actual needs.Due to the use of differential privacy mechanism,the allocation of privacy budget directly affects the performance of the scheme.This thesis proposes a new method of total allocation of privacy budget and verifies its effectiveness through experiments.In addition,this thesis theoretically proves that our scheme satisfies ?-differential privacy,and can effectively resist background knowledge attacks,while the worker's location privacy has theoretical guarantee.Finally,this thesis conducts experiments in real data sets.Experiments show that our solution can maintain a high task acceptance rate,which can effectively reduce the distance of workers accepting task paths and reduce task system overhead.This will help to fully mobilize the enthusiasm of the workers,and play a key role in promoting the more common application of SC.(2)This thesis proposes a spatial crowdsourcing worker location privacy protection scheme based on temporal correlation.Some existing schemes achieve worker location privacy protection by introducing a trusted third party.However,in some special cases,the third party is often not fully trusted,so this thesis proposes a location privacy protection scheme for SC workers based on time correlation.This thesis construct Markov model by changing the position of workers under continuous time stamp.The workers can publish the noise location directly and reduce the performance of our scheme without exposing the personal location information.It solves the problem that SC worker's location privacy protection can not effectively resist background knowledge attack in the existing technology.Because workers directly add noise to location information,worker location privacy protection can be achieved without the introduction of trusted third parties.Finally,this thesis conducts experiments in real data sets.The results show that our solution does not have a significant increase in system overhead compared to the case without privacy protection,and can maintain a high probability of task acceptance.