Evaluation of Geocoding Algorithms for Generalization-based Location Privacy

Many different smartphone applications are constantly tracking users, including individual locations and full trajectories. This information can contain sensitive information about the users that can be inferred by their whereabouts. One way to protect the user’s location is generalizing it, which consists of reducing the precision of the information so that it does not reflect the original location. Some of the existing techniques require a complex implementation that will consume computational power. This work explores the use of geocoding techniques, with information precision control, like Geohash and H3, and also proposes a new tool, based on directly generalizing decimal coordinates, named GenDec. The experiments measure data loss and distortion over a 300-point segment of a biking path. Results show that, even though all techniques indeed provide a good generalization-based protection, the jumps between scales can be very high, while GenDec can tailor the level of granularity in between scales. In addition, the location repel option of GenDec preserves a minimum noise from the original location, offering an extra protection layer. Results show that GenDed produces generalized paths that maintain path distortion and data loss, while allowing users to determine the desired level of distance, compared to H3, and in a more stable way than Geohash.