Modeling and Synthesis of GPS Trajectories for Privacy-Preserving Mobility Data Generation

This research introduces a novel methodology for generating synthetic mobility data from GPS trajectory records, addressing privacy concerns while maintaining data utility. The approach employs Functional Data Analysis (FDA) to model trajectories as multidimensional curves, capturing variations in latitude, longitude, and time-to-end. Starting from a dataset of smartphone GPS records, through spline interpolation and a two-step synthesis procedure, synthetic trajectory functions are generated, one for each original function, considering the instances with the shortest amplitude and phase distance from the reference function, performing a weighted average of their Square-Root Velocity Functions (SRVFs), and then mapping back to the original functional space. Offering scalability and adaptability to various functional inputs, this methodology advances privacy-preserving data synthesis, opening new directions for mobility data analysis.