Research On Location Privacy-preserving Task Assignment For Spatial Crowdsourcing

In recent years,the widespread prevalence of mobile smart devices has created an emerging paradigm of crowdsourcing for performing complex tasks,called Spatial Crowdsourcing.In a typical spatial crowdsourcing,when a task requester publishes a spatial task,the spatial crowdsourcing server needs to assign appropriate workers for the spatial task,and then workers equipped with mobile smart devices need to travel to the designated place to perform the task.However,to obtain the optimal task assignment effect,existing solutions generally require workers to submit their exact location to the spatial crowdsourcing server.However,given these location implies workers' sensitive information such as personal identity,health status,and living habits,there is a great potential risk of location privacy disclosure for spatial crowdsourcing workers.Therefore,optimizing the task assignment method while protecting the location privacy of workers has become a key issue to be solved in spatial crowdsourcing.Therefore,following the classical task assignment framework of location privacy protection,a task assignment scheme of location privacy protection based on differential privacy is proposed to ensure that the workers in spatial crowdsourcing can participate in task assignment effectively without exposing their location privacy.The main research contents of this paper are as follows:First,the Fisher-Jenks natural breaks-based adaptive grid partitioning algorithm,called FJAG,is proposed to protect the location privacy of workers in spatial crowdsourcing.Bernoulli random sampling technique satisfying differential privacy is used to sample the original spatial data,which solves the problem that the existing grid-based privacy spatial decomposition algorithm is difficult to deal with the large-scale and skewed spatial data in the real world.Besides,considering the traditional grid-based privacy spatial decomposition algorithm fails to effectively take advantage of the actual distribution of the original spatial data,the FisherJenks natural breaks method is used for spatial clustering of the sampled dataset to form the first layer density adaptive grid,and then the adaptive grid(AG)algorithm is used to perform two-layer adaptive partition for each first-level sub-grid.The experimental results on real large-scale spatial datasets prove the FJAG algorithm has higher query accuracy.Second,based on the regional task acceptance rate and the travel distance between workers and tasks,combined with the partial grid clipping optimization strategy,a hybrid geocast region construction(HGRC)algorithm is proposed.Experimental results based on real datasets prove that the HGRC algorithm can ensure the high success rate of task assignment and reduce the system overhead effectively.