| In recent years,with the development of big data,the Internet of Things,cloud computing,and other emerging technologies,mobile crowdsensing has become a kind of data collection platform that has attracted much attention.Compared with the traditional way of installing fixed sensors,mobile crowdsensing has the advantages of lower cost,wider range,and higher quality and control of data.Mobile crowdsensing has unique advantages and plays an important role in urban management,environmental monitoring,traffic management,noise detection,medical health and other fields.In mobile crowdsensing systems,task assignment is one of the key aspects which determines the workload and data quality of the participants.The system usually performs optimal task assignments based on participants’ information,such as users’ precise location or bids,to improve stability and efficiency.However,the collection and analysis of users’ data may jeopardize their privacy.Existing privacy-preserving approaches for task assignment focus on a single target constraint,rarely consider comprehensive protection of user privacy,and less consider privacy protection during dynamic task assignment.The main research work in this paper is as follows:1.A multi-objective optimized task assignment differential privacy protection method is proposed by considering the mobile swarm intelligence-aware user privacy protection with multiple constraints in a comprehensive manner.Without any trusted third party,the multiobjective optimized task assignment differential privacy protection scheme enables participants to obfuscate their sensitive data using differential privacy techniques.Based on the game theory using non-dominated ranking genetic algorithm-II to optimize multiple objectives to minimize the expected travel distance of selected workers and the cost of the task publisher.The experimental results show that the scheme in this paper outperforms the Laplace’s confusion scheme on real-world data,with a 25.6% reduction in the average distance traveled and a 63.6%reduction in the average expenditure cost.Compared with the single-objective optimization scheme,the average moving distance is reduced by 41.3% and the average expenditure cost is reduced by 64%.2.Based on the multi-objective optimization task assignment differential privacy protection model,we further study the online task assignment privacy protection model for streaming data by constructing dynamic queues based on the formalization of data streams for the real-time and dynamic characteristics of mobile swarm intelligence perception systems.Based on the Lyapunov optimization technique,we formally construct a dynamic queue of tasks and study the online task assignment method of the mobile group intelligence awareness platform.By introducing a virtual queue,the online system is converted into a dynamic queue system,and the long-term optimization objective and constraint problem is modeled as an optimization problem that minimizes the perceived cost to achieve a reasonable online task assignment strategy.Meanwhile,for the risk of user privacy leakage,user information is abstracted into data streams,and the data streams are perturbed using differential privacy techniques based on the exponential mechanism to achieve privacy protection of user data during dynamic task allocation.The experimental results show that although the average queue capacity of this paper’s scheme will be slightly higher than the lowest queue backlog task allocation scheme,it can still maintain long-term stability and protect user privacy. |
