A Parallel Algorithm For Anonymizing Large-scale Trajectory Data

Abstract

With the proliferation of location-based services enabled by a large number of mobile devices and applications, the quantity of location data, such as trajectories collected by service providers, is gigantic. If these datasets could be published, then they would be valuable assets to various service providers to explore business opportunities, to study commuter behavior for better transport management, which in turn benefits the general public for day-today commuting. However, there are two major concerns that considerably limit the availability and the usage of these trajectory datasets. The first is the threat to individual privacy, as users' trajectories may be misused to discover sensitive information, such as home locations, their children's school locations, or social information like habits or relationships. The other concern is the ability to analyze the exabytes of location data in a timely manner. Although there have been trajectory anonymization approaches proposed in the past to mitigate privacy concerns. None of these prior works address the scalability issue, since it is a newly occurring problem brought by the significantly increasing adoption of location-based services. In this article, we conquer these two challenges by designing a novel parallel trajectory anonymization algorithm that achieves scalability, strong privacy protection, and high utility rate of the anonymized trajectory datasets. We have conducted extensive experiments using MapReduce and Spark on real maps with different topologies, and our results prove both effectiveness and efficiency when compared with the centralized approaches. CCS Concepts: • Security and privacy → Social aspects of security and privacy; Privacy protections; • Computing methodologies → MapReduce algorithms; Massively parallel algorithms;

A systematic overview on methods to protect sensitive data provided for various analyses

“…Even methods in the f (federated) category could be problematic in such cases, as the data quality and differences between the local data repositories cannot be assessed adequately. Almost all methods need extensions to tackle challenges such as high-dimensional data [65], anonymization of event-history data [39], anonymization of trajectory/mobility data [85], or time-varying sensitive features.…”