Application of a predictive method to protect privacy of mobility data
Abstract
Users of geo-localized applications on mobile devices need protection to avoid threats to their privacy. Such protection should vary in time, to cope with the dynamical nature of mobility data. We present a method to protect the privacy of users of location-based services, based on Model Predictive Control techniques. We employ three different predictors for future movements: an exact predictor, which serves as the baseline for the best expected performance, and two additional predictors allowing for online implementation. One of these predictors assumes the user is moving in a way that minimizes privacy, while the other is a linear predictor. The method has been applied to two datasets, Privamov and Cabspotting, which contain mobility data collected from real users when using a mobile device. The method demonstrated an improvement in privacy compared to a state-of-the-art mechanism by approximately 12% increase for Privamov users and 5% for Cabspotting users, while maintaining the same level of utility.
Origin | Files produced by the author(s) |
---|---|
licence |