Users collaborative mix-zone to resist the query content and time interval correlation attacks

Abstract

In location-based services of continuous query, it is easier than snapshot to confirm whether a location belongs to a particular user, because sole location can be composed into a trajectory by profile correlation. In order to cut off the correlation and disturb the sub-trajectory, an un-detective region called mix-zone was proposed. However, at the time of this writing, the existing algorithms of this type mainly focus on the profiles of ID, passing time, transition probability, mobility patterns as well as road characteristics. In addition, there is still no standard way of coping with attacks of correlating each location by mining out query content and time interval from the sub-trajectory. To cope with such types of attack, users have to generalize their query contents and time intervals similarity. Hence, this paper first provided an attack model to simulate the adversary correlating the real location with a higher probability of query content and time interval similarity. Then a user collaboration mix-zone (CoMix) that can generalize these two types of profiles is proposed, so as to achieve location privacy. In CoMix, each user shares the common profile set to lowering the probability of success opponents to get the actual position through the correlation of location. Thirdly, entropy is utilized to measure the level of privacy preservation. At last, this paper further verifies the effectiveness and efficiency of the proposed algorithm by experimental evaluations.