On the effectiveness of differential privacy to continuous queries

Location-based services have a wide range of applications; more recent among those include monitoring virus spread and disaster management. Nevertheless, they also open up new vulnerabilities in terms of the user’s location privacy. Differential privacy has been widely accepted to provide location privacy as it provides a provable privacy guarantee. In this paper, we show that a differential privacy mechanism designed for individual queries is not effective when applied repeatedly to queries from correlated locations. We call such location-based queries from adjacent locations of a user as continuous queries. We quantify the reduction in privacy level of differential privacy when applied to continuous queries. The problem is solved from an adversarial viewpoint, given the perturbed location of trace size n, estimate the actual location. Assuming \(\epsilon \) $ϵ$ is the privacy level for applying noise independently to a location, we prove that the privacy level is reduced to \(n\epsilon \) $nϵ$ , when applied to n consecutive locations. We propose a privacy preserving mechanism and show that it handles continuous queries better than standard differential privacy mechanisms. The results are validated empirically using a real dataset.