Impact of prior knowledge on privacy leakage in trajectory data publishing

Abstract

The rapid growth in the usage of location-based services has resulted in extensive research on users’ trajectory data publishing. But, a key concern here is a potential breach of user privacy through various linkage attacks by an efficient adversary. There exist a few privacy preservation methods to defend against either single or combination of linkage attacks, namely Identity linkage attack, Attribute linkage attack, and Similarity attack. However, the Correlated-records linkage attack has not been studied in any previous privacy preservation methods, and there is no privacy preservation method to address all the above four linkage attacks. In this paper, a novel anonymization method is proposed to provide the privacy guarantee to users against all the four linkage attacks. The proposed method consists of two phases, namely virtualization and suppression. The virtualization method works as a replacement mechanism for the sensitive attribute and the suppression method works as anonymization mechanism for users trajectories, in order to anonymize the trajectory datasets for preserving users’ privacy from the above four linkage attacks. To validate the efficiency of the proposed method, it is also compared with existing methods, namely KCL-L, KCL-G and KCL-PPTD, considering both synthetic and real-time datasets. The experimental results exhibit that the proposed approach results in better performance with a significant reduction in the information loss when compared to other states of the art methods.