Traffic-sensitive navigation systems in vehicular networks help drivers avoid traffic jams by providing several realtime navigation routes. However, drivers still encounter privacy concerns because their sensitive locations, i.e., their start point and endpoint, are submitted to an honest-but-curious navigation service provider (NSP). Previous privacy-preserving studies exhibit serious deficiencies under similar queries: if a driver makes several similar queries, i.e., periodically makes requests for the same start point and endpoint to the NSP, these requests will eventually reveal the areas of the two points as well as the route. In this article, we present a novel privacy-preserving navigation scheme PiSim, which supports similar queries in navigation services. Intuitively, we transform the typical navigation approach into a traffic congestion querying approach. Instead of sending two locations to the NSP and awaiting a navigation route, drivers query the traffic congestion along the navigation route. Specifically, PiSim is characterized by extending anonymous authentication, facilitating privacy-preserving multi-keyword fuzzy search, and constructing weighted proximity graphs. Our scheme protects location privacy and route privacy, and defends against multiple requesting, spurious reporting, and collusion attacks from malicious drivers. Finally, a detailed analysis confirms the privacy and security properties of PiSim. Extensive experiments are conducted to demonstrate the feasibility, performance, and privacy protection level.

@en

Ride-Hailing Service (RHS) has motivated the rise of innovative transportation services. It enables riders to hail a cab or private vehicle at the roadside by sending a ride request to the Ride-Hailing Service Provider (RHSP). Such a request collects rider’s real-time locations, which incur serious privacy concerns for riders. While there are many location privacy-preserving mechanisms in the literature, few of them consider mobility patterns or location semantics in RHS. In this work, we propose a pick-up location recommendation scheme with location indistinguishability and semantic indistinguishability for RHS. Specifically, we give formal definitions of location indistinguishability and semantic indistinguishability. We model the rider mobility as a time-dependent first-order Markov chain and generates a rider’s mobility profile. Next, it calculates the geographic similarity between riders by using the Mallows distance and classifies them into different geographic groups. To comprehend the semantics of a location, it extracts such information through user-generated content from two popular social networks and obtains the semantic representations of locations. Cosine similarity and unified hypergraph are used to compute the semantic similarities between locations. Finally, it outputs a set of recommended pick-up locations. To evaluate the performance, we build our mobility model over the real-world dataset GeoLife, analyze the computational costs of a rider, show the utility, and implement it on an Android smartphone. The experimental results show that it costs less than 0.12 ms to recommend 10 pick-up locations within 500 m of walking distance.@en

Location-based services (LBSs) are widely deployed in commercial services. These services always depend on a service provider, e.g., a cloud server, to store the enormous amounts of geospatial data and to process various queries. For example, a Yelp user can retrieve a list of recommended cafés by submitting her/his current location to the service provider. While LBSs offer tremendous benefits, it is vital to safeguard users’ privacy against untrusted service providers. However, no prior secure k nearest neighbor query processing schemes satisfy the three security requirements of one-time, oblivious, and unlinkable. In particular, we are concerned with the problem of item exclusion: how to match one data query with each item on the cloud no more than once in an oblivious and unlinkable manner. In this paper, we propose the first secure k nearest neighbor query processing scheme, Obaq, that satisfies the above requirements. Obaq first introduces an item identifier into an existing secure k nearest neighbor query processing scheme. Each data owner inserts an item identifier and her/his location information into a secure index, and each data user transfers the identifier of a previously received data item and location information into a specific range. Then, Obaq excludes corresponding items via privacy-preserving range querying. We define strong index privacy and strong token privacy and formally prove the security of Obaq in the random oracle model. We further evaluate the performance of Obaq using a prototype and a real-world dataset. The experimental results show that Obaq is highly efficient and practical in terms of computational cost, communication overhead, and response delay.

@en