Geospatial streams publish with differential privacy

Abstract

Continuous releasing geospatial data is benefiting numerous areas, such as information push service, traffic scheduling and task assignment in crowdsourcing, etc. This kind of data is generated by people using positioning service in daily life, from which much sensitive information can be derived. Differential privacy is a strong theoretical and practical tool to provide protection; it has already been used on streams composing by datasets with fixed attributes. However, there is limited work on geospatial stream releasing with dynamic scopes for the requirement of accurate query. In this paper, aiming at achieving privacy protection of real-time geospatial synopsis with high utility, we introduce a method, called Realtime Geospatial Publish (RGP), which adopts differential privacy to geospatial stream with a new structure k-memo. We prove the privacy and utility of RGP theoretically and show the improvement of utility by experimental comparison with existing approaches on real datasets.