Privacy Preserved Data Publishing of Moving Object Trajectories

The pervasiveness of inexpensive and sophisticated global positioning system enabled mobile devices within the location-based systems has led the authorities to collect moving traces of objects called spatio-temporal trajectories in an unprecedented manner. This has created a unique opportunity for the researchers and other stakeholders to dig valuable insights from published trajectory data and has been used to manage or create new applications in the fields like urban planning, transportation, epidemiology etc. But the tracking of users and publishing user trajectories without consent of the people is a privacy encroachment and it has been protected by the law around the world. This activity may lead to uncovering the valuable sensitive details about the individuals such as sexual preferences, health details, religious customs etc. into the hands of adversaries. However, publication of datasets by the data owners is essential for the developmental activities. So, every researcher or data owner has aimed to publish the trajectory data by balancing the trade-off between data utility and individual privacy. This urges the need for anonymization before publishing the trajectory.

In this book, instead of anonymizing the whole trajectory, we focused on some halting points or stay locations on the trajectory as most sensitive to the user and anonymized them to provide privacy. In the initial work, we have applied a generalization-based anonymization for the trajectory segment with sensitive halting points in an area zone containing other non-sensitive halting points. In the whole work, we used Haversine instead of Euclidean measure for the distance-based calculations, since the Haversine considers the spherical nature of the earth. In the next work, we derived a unique sensitive stay location finder function to extract sensitive stay locations from the user trajectories by considering the various parameters derived from trajectories and applied generalization-based anonymization on these to reduce the information loss. The succeeding work suggests a perfect blend of existing generalization techniques with Places of Interest along with a novel temporal perturbation technique for the anonymization of sensitive stay locations. Finally, we proposed a personal privacy protection model through the anonymization of most sensitive stay locations, like home, work places etc. for the collected Covid-19 pandemic data and the data thus anonymized can be used by the researchers and health authorities to act brilliantly if similar situations arise in future. All these works creates Haversine distance measured MBR stay zones, over the most sensitive stay locations with similar Places of Interest to provide anonymity and also, it prevents the adversary from getting known to the exact information about the sensitive stay locations.

The experiments and evaluations with real-world datasets prove that our approaches reduce unnecessary anonymization of trajectory segments or even trajectories and provide high data utility, less information loss and greater privacy for the users during the trajectory publication. Our approaches also prevent adversaries from deriving inferences about the sensitive stay locations and planning attacks based on this. The generalization cum perturbation approach is especially useful for preventing various types of attacks like user profiling, behavioral and contextual attacks or even different types of attacks against VVIPs (Very Very Important Person).