Gao, J., Lu, Q. L. & Cai, M. (2020). Quantifying privacy vulnerability under linkage attack across multi-source individual mobility data. In 99th Transportation Research Board (TRB) Annual Meeting.

With the advances in detector and sensor technologies, identity detection-based intelligent transportation systems—such as license plate recognition (LPR) system and parking electronic toll collection (ETC) system—have been widely deployed in urban transportation, generating large quantities of multi-source individual-based mobility data set (e.g., LPR data and parking data). Given the high frequency, precision and wide coverage, these individual-based mobility data can be used in many transportation research areas, such as transportation planning, traffic prediction and individual mobility pattern profiling. With the increasing demand for publishing and sharing these individual-based data sets to researchers and practitioners, the privacy issue of data publishing has been a major concern since true identities of individuals can be revealed by linkage attack. In this paper, we quantitatively measure the privacy disclosure risk caused by linkage attack across multi-source individual-based mobility data sets. Taking an example of LPR data and parking data, a traffic-knowledge-driven adversary model is proposed for linkage attack conducting among LPR data and parking data. Two common modes of LPR data publishing are examined and two quantitative criteria are introduced to present the risk of privacy leakage under linkage attack. The experimental results demonstrate that anonymized individual still under high risk of being linked successfully (71.63% under mode 1 and 36.55% under mode 2). This study serves as a wake-up call for relevant agencies and data owners about the privacy vulnerability caused by linkage attack across multi-source individual-based mobility data.