Research On Differential Privacy Protection Method For Location Recommendation

With the development of mobile location technology and mobile Internet,LBS(Location Based Service)have been applied widely.Especially in the era when the value of data is becoming more and more important,LBS system has a strong desire to analyze and mine user's location information in depth so as to provide a better recommendation service about location and the surrounding information with the more accurate and more personalized ability.However,there are some risks of location privacy in the process of data analysis and mining.If the risks are ignored,they will endanger user's personal privacy such as health condition,income level and other sensitive information.Therefore,how to protect the user's location privacy without affecting the quality of location services has become an important research topic in the current.Traditional methods for location privacy protection cannot be implemented effectively because of the defect of failing to quantify and an assumption based on limited background knowledge.But,differential privacy has been paid more concern for it solves the defects of the traditional privacy protection method,and has been become a research hot spot in the field of privacy protection in recent years.Based on this,a new location privacy protection method based on differential privacy for location recommendation service is proposed in this thesis.The method which mainly uses strict mathematical definition of differential privacy allocates reasonably the differential privacy budget based on path prefix structure.In addition,the prposed method can handle the tilting problem relaterd to privacy budget allocation based on geometrical principle.The method make the location data after adding Laplace noise can better keep the regularity of user's original trajectory and the feature of location visited frequency,thus to reduce the impact on recommendation effect.At the same time,in order to improve the authenticity and availability of location data after adding Laplace noise,it is necessary for the location data containing the noise to meet the consistency constraint of path prefix structure.In terms of the constraint conditions of descending order and integer,the thesis also propose a consistency post-processing method respectively based on the way of quadratic programming problem and the principle of error minimization.In the end,the thesis has analyzed the privacy of the proposed differential privacy method from the theoretical perspective.Effectiveness experiment and error experiments are made to analyze the proposed location privacy protection method based on differential privacy.The experimental results show that the proposed method can effectively protect user's location privacy and effectively reduce the influence of differentia l privacy noise on recommendation effect in the location recommendation,thus to make the recommendation has a good effect in the differential privacy protection.