Enhancing Privacy in Ride-Sharing Applications Through POIs Selection

The problem of data privacy preservation is of central importance in ride-sharing applications, because in order to efficiently match passengers with vehicles, these services rely on exact location information. Yet, transportation and location data can reveal personal habits, preferences and behaviors, and users may prefer not to share their exact location. Masking location data in order to avoid the identification of users in case of data leakage, and/or misusage would help protect user privacy, but could also lead to poorer system performance, in terms of efficiency and quality of service as perceived by users.In this paper, we compare classic data masking techniques, namely obfuscation, k-anonymity, and l-diversity, applied to users' location data, before sending it to a carpooling system. While the first two techniques use randomly generated points to mask the actual location, l-diversity uses actual points of interest, having the additional benefit of ensuring that the disclosed location is always an accessible and safe pickup or drop-off location. Given that users in a real ride-sharing system could choose to protect or not protect their location data when using the system, we also evaluate the effect of privacy preservation penetration rate, by varying the percentage of users choosing to have their location data protected. The results show that l-diversity performance is better than the others' even when the privacy penetration rate is high, suggesting that this technique has the potential to meet both users' and system's needs, and thus being a better option to provide privacy within carpooling systems.