Research On Differential Privacy Protection Method For Service Rating

With the continuous innovation and progress of the Internet,while users enjoy various types of services provided by service providers(SP),service-related data is also increasing.These data are collected and analyzed by SP to obtain useful information for their services.Service rating is a common way to obtain such data.After users participate in the service,rating data of different dimensions is given,which is collected by the SP to assist in improving the service.However,as users' privacy awareness continues to increase,they no longer trust others,and they are only willing to share data when private data is properly disinfected before leaving their devices.Therefore,facing the data collection behavior of untrusted SP,the local differential privacy(LDP)model is often used to protect service rating data,but the existing LDP protocol has low estimation accuracy,and the SP-side computational cost is too high and cannot meet the user's different privacy protection needs and other issues.In view of the above problems,the main research work of this paper is as follows:(1)Aiming at the problems of low accuracy of service access frequency estimation and excessive computational cost of SP-side in the existing LDP protocol,a solution for surveying service access frequency in local settings is proposed to achieve privacy of user access information on the local side Statistics and service access frequency statistics on the SP-side.This solution first introduces an edge computing model to design a service access data collection framework,and then uses one-hot encoding to design a data perturbation algorithm to avoid algorithm errors due to hash conflicts in the LDP protocol based on hash encoding and simplify the randomization step to improve the accuracy of frequency estimation,and then upload the disturbance data to the edge layer and calculate the intermediate results required for frequency estimation to reduce the computational cost of the SP-side.Finally,through theoretical analysis,it shows the advantage of this scheme in terms of SP-side computational cost compared with other data collection schemes.After experimental verification,the designed perturbation algorithm has higher frequency estimation accuracy than the representative method.(2)Aiming at the problem that the existing LDP protocol's lack of recognition of user privacy preferences leads to a decrease in the accuracy of aggregated rating data,a rating data aggregation scheme based on user privacy preferences is proposed.The scheme uses the history of user privacy preferences and service ratings to quantify the user's rating sensitivity,divides the user's privacy protection level according to the obtained sensitivity to implement data perturbation,and then designs an LDP protocol based on key-value encoding to perturb the overall rating The data can be used to estimate the frequency of the service and the mean value of the score.Finally,the privacy and performance analysis of the scheme is carried out and verified through experiments.The scheme meets the different privacy protection needs of users and has higher data practicability than representative methods.