Continuous Location Statistics Sharing Based On Local Differential Privacy

Valuable knowledge has been created through continuous sharing of location statistical data,and many third parties who are interested could obtain beneficial information through analyses of the data.In such applications,what should be paid attention the most is that data should be shared without compromising user privacy.Differential privacy has become a feasible and reliable technology for the release of privacy statistics.In the process of privacy protection,a lot of data processing tasks are concentrated on the server,and the attacker might access the real data by attacking the server.However,local differential privacy could solve such attacks very well,that is to avoid such risks by obfuscating each user's data before these data reach the server for statistical analysis.First,a kind of location data collection method suitable for local differential privacy is proposed to address the privacy disclosure in the process of collecting location data.In order to be applicable to local differential privacy,a multi-stage random response mechanism is adopted in the process of location data collection.Besides,the estimation method of regional location density is set in the application process of different scenes for estimating regional density.The definition is to analyze the relationship between location data and information collecting nodes within a certain time interval,in order to determine the location density within the corresponding associated area.This method protects user data through local differential privacy in applications,at the same time,unreal data could be used to estimate the area density under the premise of not exposing the user's real data.In this paper,a shared algorithm of location information statistics combining local differential privacy and ?-event privacy is proposed.The algorithm first calculates the similarity result and performs a random response mechanism to it,and makes a decision on whether to add Laplacian noise on this basis,then it compares the estimated privacy budget of the forward sliding window projection and the remaining privacy budget of the backward sliding window projection,and takes the minimum value between the two to ensure that the privacy budget can be consumed as little as possible.The proposed solution guarantees users' privacy when publishing statistical data continuously on unlimited data streams,and is of strong privacy protection capabilities.Finally,it can be known that the data obtained based on the local differential privacy data collection can be estimated to reflect the real situation through experiments with simulated distribution and real data;the availability of privacy publishing algorithm based on local differential privacy is not inferior to the mainstream privacy budget absorption algorithms and privacy budget distribution algorithms.