Practical Privacy-Preservation Techniques for Location-Based Services

Speaker Bio

Abdur Rahman Bin Shahid is a Ph.D. candidate at Florida International University’s School of Computing and Information Sciences under the supervision of Dr. Niki Pissinou. He received his B.Sc. in Computer Science and Engineering in December 2011 from Chittagong University of Engineering and Technology, Bangladesh.  He worked as a Software Engineer at Samsung Bangladesh R&D Center for two years in his country before joining the SCIS Ph.D. program. His research interests include location-based applications, privacy, and security. His current research focuses on privacy preservation in location-based services.

Description

Location-Based Services (LBS) offer users a vast array of services such as Point-of-Interest (POI) searching, check-ins and geotagged photo sharing in Location-Based Social Networks (LBSN). However, these services come at a price; they offer service in exchange for user’s precise spatial locations. The disclosed spatiotemporal information could reveal sensitive information, such as interests, health status, frequently visited places, relationships, and raise serious privacy concerns in LBS. The goal of location privacy-preservation in LBS has attracted significant attention from researchers and has delivered a variety of Location Privacy-Preserving Mechanisms (LPPMs), including k-anonymity, dummies, spatial obfuscation, and cryptography. Existing LPPMs exhibit different weaknesses against a wide range of attacks, derived from a plethora of LBS usage data. In this dissertation, we focus on four limitations of existing mechanisms. First, the generalized modeling of privacy is ineffective against an attacker with knowledge of an individual’s LBS usage. Second, the widely used random mobility model does not reflect the real mobility of users, and thus cannot guarantee reliable privacy. Third, most research efforts concentrate on POI search, leaving the impact of other services (e.g., geotagged photos) on location privacy to be analyzed. Fourth, there is a lack of privacy preservation for a group of users which would otherwise ensure personalized privacy and quality of services. In this context, we aim to devise efficient mechanisms to improve these limitations.

At the outset, we propose a Location Privacy-Preserving Mechanism, called KLAP. KLAP utilizes the advantages of knowledge of the map, location semantics and historical queries to personalize a user’s privacy. These advantages make KLAP resilient against map matching, location homogeneity, probability distribution, travel-time information, and personal context linking attacks. We implement KLAP with a real dataset from Foursquare to analyze its efficiency. Afterward, we take a step toward understanding which LBS usage information, besides query and check-in, influence location privacy. Specifically, our analysis reveals that distribution of historical photos over locations, shared in LBSN, can be used to design a location inference model. We implement this inference model on existing LPPMs to quantify their vulnerability. Based on the findings, we introduce a LPPM, called photo-check, for check-in and photo sharing in LBSN.

Furthermore, we propose to extend our work in two directions. First, we will study the Markov chain mobility model to preserve location privacy in spatial, temporal, and semantic domains for both frequent and continuous LBS utilizations. Second, we aim to explore privacy preservation techniques for a group of users without compromising personalized privacy and efficiency. The successful outcomes of this dissertation will enhance location privacy in practical location-based applications.