Privacy-Preserving Trajectory Publication against Parking Point Attacks

GPS data is becoming more and more popular due to massive usage of global positioning systems, other location-based devices and services. However, publishing original GPS data to the public or a third party for data mining and analysis could cause serious privacy issues. In this paper, we perform a study on taxi GPS data and identify a new type of attack called a parking points attack. In a parking points attack, an adversary utilizes parking habits of taxi drivers to re-identify related victims in a published taxi trajectory dataset. To against such attacks, we introduce the concept of spatial-temporal tunnel to swap sub-trajectories of taxis. Using real GPS trajectories from more than 12,000 taxis, we demonstrate that the proposed approach effectively limits parking point attacks. As a result, more than 55% of trajectories can be re-identified at a probability of 1 in original trajectory dataset, but only 8.6% of trajectories can be re-identified in our swapped trajectory dataset.