Real-time sharing of privacy protected location data in road networks

Abstract

Location-based applications like nearby information services (e.g., nding the nearest bus station or a hospital), tra c monitoring, urban planning and transport management facilitate the development of smart cities and improve the quality of life for citizens. A user's location is a sensitive data and can reveal private information about the user's health, habit and preferences. Due to privacy concerns, people may hesitate to share their locations and prohibit the growth of location-based services and analysis. In this thesis, we develop a novel approach to share privacy protected location data with others in real time. Our approach does not need to trust any party including a centralized server or peers in a distributed setting for protecting location privacy. Researchers have developed techniques for sharing a user's locations in a privacy preserving manner in the Euclidean space and road networks. However, none of these approaches can ensure both user privacy and data utility in real time. We identify the possible privacy threats in the literature and develop solutions to overcome the privacy attacks. In our approach, a user reveals a cloaked region (i.e., a region that includes her current location) instead of an actual location if the disclosure of the user's actual location enables others to infer a user's visit to a sensitive place. Existing approaches fail to protect a user's privacy for not considering upcoming sensitive locations in advance. We develop a technique to precompute the warning zone, i.e., the re ned area where the disclosure of a user's actual location may enable adversaries to identify the user's sensitive locations. Warning zones also enable users to reduce the frequency of not sharing locations for privacy reasons and thereby improve the accuracy and utility of shared locations. In addition, we develop an algorithm to compute a user's cloaked regions using the pre computed warning zones with reduced processing time. We evaluate our proposed approach using a real dataset and compare our algorithm with the most recent stateof- the-art technique in road-networks, in terms of privacy, data utility and computational overhead.