| The wide application and development of mobile intelligent terminals and wireless sensing devices provide great convenience for people’s lives,work and learning.The crowdsourcing approach of using personal mobile devices and wireless network infrastructures to collect and analyze ultra-large scale sensory data is also becoming more familiar to users,however,this raises many concerns about personal privacy.Most of the existing privacy-preserving approaches start from the perspective of data distribution,such as k-anonymity,randomized perturbation,blocking,etc.,while this paper focuses more on thinking about privacy preservation from the perspective of data collection.The negative investigation approach,inspired by the idea that immune cells in the biological immune system recognize and match cells through a "self-non-self" mechanism,was proposed by Esponda et al.and has received much attention because of its simplicity and efficiency.However,the existing negative survey methods suffer from problems such as negative values generated by the reconstruction results and low computational efficiency.Therefore,in this paper,two negative survey methods based on negative quadtrees are proposed to address the above problems.The specific research work is as follows.(1)A study of negative quadratic tree method based on linear programming.In this paper,we abstract the estimation of positive survey results from negative survey results as a singleobjective optimization problem and solve this linear programming problem by interior point method.Users upload their location data anonymously locally,and the server reconstructs the original user data distribution according to the proposed NQT-LP algorithm based on the anonymized data uploaded by users.Based on the results of the simulation experiments,it can be seen that the method neither generates negative values nor incurs excessive computational costs compared to the existing methods of evaluating positive survey results from negative survey results.(2)A study of negative quadtree methods based on difference minimization optimization.Difference minimization optimization of a negative quadtree method based on linear programming for the problem of positional correlation attack.When the attacker has collected enough history of the user,he can obtain the actual location of the user by comparing the negative locations collected in adjacent time periods and continuously reducing the number of possible locations.Therefore,the method proposed in this paper tries to reuse the elements of the previous negative vector.Then assuming that the user does not move very far,most of the negative vector elements can be reused.It is experimentally demonstrated that the NQT-LP method optimized by difference minimization can always maintain a high logarithm of possible locations and effectively resist location-based correlation attacks.(3)The proposed two negative quadtree methods are applied to the monitoring of traffic road density and the statistics of crowd flow in the area of interest respectively,extending the application areas of the methods.The experimental results demonstrate that the proposed method has good privacy protection in both simulated and real datasets.Users can choose the appropriate parameter values to adjust the privacy protection level according to their privacy protection needs. |
