Privacy and spatial pattern preservation in masked GPS trajectory data

Abstract

ABSTRACTPersonal trajectory data are increasingly collected for a variety of academic and recreational pursuits. As access to location data widens and locations are linked to other information repositories, individuals become increasingly vulnerable to identification. The quality and precision of spatially linked attributes are essential to accurate analysis; yet, there is a trade-off between privacy and geographic data resolution. Obfuscation of point data, or masking, is a solution that aims to protect privacy and maximize preservation of spatial pattern. Trajectory data, with multiple locations recorded for an entity over time, is a strong personal identifier. This study explores the balance between privacy and spatial pattern resulting from two methods of obfuscation for personal GPS data: grid masking and random perturbation. These methods are applied to travel survey GPS data in the greater metropolitan regions of Chicago and Atlanta. The rate of pattern correlation between the original and masked data sets declines as the distance thresholds for masking increase. Grid masking at the 250-m threshold preserves route anonymity better than other methods and distance thresholds tested, but preserves spatial pattern least. This study also finds via linear regression that median trip speed and road density are significant predictors of trip anonymity.