Privacy Risk Evaluation of Human Mobility Data for Urban Transportation Planning

Abstract

This project examined the variations in re-identification risks of mobility trace data in different urban areas, characterized by residential population density, percentage of residential land use, and per capita income, and in different population segments, characterized by race, gender, and household income. The project used the 2017 Puget Sound Regional Travel Survey and estimated the uniqueness of the trip origins and destinations by using the method of k-anonymity. This project found that 42 percent of travelers could be re-identified by one trip origin or destination point aggregated at the census block group level and the one-hour time interval. This confirmed previous findings that mobility traces are highly unique. This project further estimated the associations between the built environment and sociodemographic variables and the k-values that measure the uniqueness of mobility traces in a data set. The results showed that trips to or from census block groups with a lower per capita income, higher residential population density, or higher percentage of residential land use were more likely to have a higher level of re-identifiability. Similarly, travelers whose mobility traces were more unique than others tended to have higher percentages of male, non-white, and lower income populations. The findings help to optimize the algorithmic solution to minimize the privacy risks and detect and mitigate algorithmic biases in current data practices.