Our research has been continuously funded by Research Grants Council (RGC) of Hong Kong, as well as supported by the Hong Kong Scholars Program, Croucher Foundation, K. C. Wong Education Foundation, and Tin Ka Ping Foundation. In the past years, our research group has received over $30 million of funding for 40+ research grants, including 29 GRFs (PI), 2 ECS (PI), 2 ITFs (PI), 2 CRFs (Co-PI), and 2 RIFs (Co-PI).

Owing to recent advances in data-as-a-service (DaaS) and cloud computing, cloud-based analytical query services have become widely available. For example, an aggregate query in personal genomics analysis aims to identify frequent gene patterns associated with some genetic diseases; an association rule query in market basket analysis seeks to find associations between customer purchases. Analytical query results can be used to inform critical decision making and derive valuable business intelligence. However, as the cloud service provider, which is often a third party, might not be fully trustworthy, it is imperative to authenticate the integrity of analytical results. This project investigates novel query authentication techniques for set-valued data.

The research on graph data has been rejuvenated with a large range of recent real-world applications, ranging from social networks, information networks, and co-purchase networks to biological and chemical databases. Due to the volume of data graphs and a lack of IT expertise, hosting efficient graph query services has been a technically challenging task for the owners of graph data. Hence, they may prefer to outsource these services to third-party service providers (SPs) for scalability, elasticity and performance. Some companies (especially, in the pharmaceutical domain) have been supporting structural search services. SPs may not always be trusted. Security, including the confidentiality of data (and of queries), has been recognized as one of the critical attributes of quality of service (QoS). In this project, we investigate a novel unified framework for privacy-preserving linear algebra that is generic enough to implement a wide range of graph query algorithms.

The graph model has been widely used to model complex relationships among entities. Examples include DBLP (citation relationships), Facebook (friend relationships) and the whole internet (link relationships). Many algorithms and systems are designed to work with these graphs. However, these graphs are growing more rapidly than the capacity of machines. The rate of growth makes it harder to work on these graphs as the main memory cannot hold the whole graphs and without careful design, a graph algorithm usually suffers from high latency when the graph has to be kept on disk. In this project, we focus on I/O efficient graph algorithms in which the size of the graph exceeds the size of the main memory and we will investigate graph storage, I/O efficient graph algorithms and system design.

The usability issue is a pressing issue of graph repositories, because graph schemas may be too loose to be useful for query formulation; the queries can be too complex to compose manually; or the user may simply fall victim to human error. To address this issue, this project investigates useful feedback to guide users to retrieve and/or explore graph repositories. In this project, we formalize subgraph query feedback into two novel and fundamental queries. Our pioneering effort on a graphical user interface (GUI) of a graph platform has been a natural means to provide such feedback to facilitate graph exploration.

With the deep penetration of smartphones and geo-locating devices, ridesharing is envisioned as a promising solution to mitigate transportation-related problems such as congestion and air pollution for metropolitan cities like Hong Kong. However, social discomfort and safety concerns about traveling with strangers have been notable barriers to ridesharing. To overcome these barriers, this project proposes social-aware ridesharing services, whereby participants' social connections are considered besides spatio-temporal proximity in assigning them to shared rides.

The convergence of sensor technologies, crowdsourcing, and cloud computing has given rise to a brand-new paradigm for cloud-assisted multi-source data collection, where data is collected by physically distributed collectors and outsourced to the cloud to provide query services. While this model is appealing in terms of cost, performance and flexibility, it raises the issue of query integrity. This project aims to investigate query authentication techniques to empower clients to verify the integrity of query services in cloud-assisted multi-source data collection environments.

The recent rapid rise of information networks in various application domains has generated enormous network data, which are typically represented as graphs with nodes corresponding to a set of individuals and edges corresponding to connections between them. While network data have flourished a wide spectrum of data analysis tasks, it makes individual privacy more subject to intrusions than ever before. This project proposes to investigate novel practical techniques for non-interactive network data publication under differential privacy.

A geo-proximity service in location-based social networks alerts a mobile user when any of his/her friends is in the geographical vicinity, so as to enrich social activities such as collaborative working and information sharing. To realize such services, existing systems collect location information from mobile users for proximity computation, which raises serious privacy concerns. This project aims to develop more sophisticated location update and query techniques that support these geo-proximity services while preserving the location privacy of mobile users.

In location-based services, there has been a growing necessity against location cheating and for location trustworthiness. In this project, we propose "spatio-temporal attestation", where a mobile user testifies or attests to a service provider the genuineness of his/her input location against some spatio-temporal predicate, such as "being in a specific region during a time period". The major challenge is the need of protecting location privacy of mobile users against the service provider during the attestation process.

In data oursourcing, it is critical for a query client to be able to authenticate the result of a query from the outsourced server, in terms of both soundness and completeness. However, existing works assume that during the authentication process, the client can always be trusted and entitled to receive data values on the querying attribute(s), even if they are not the results. This severely jeopardizes the privacy of the data owner. In this project, we study authenticationfor such a privacy-conscious query model where the querying attribute(s) are unavailable to the client.

Graph/Network data model has been a powerful tool to model complicated structures such as social networks, network traffic, biological databases and XML documents, among many others. A typical task on graph data is to retrieve substructures embedded. The pruning power of indexes is evidently crucial to the overall performance. However, data graphs are often changing or evolving, which may affect the indexes' pruning powers. In this project, we propose to adjust the indexes adaptively in response to query workloads.

Owing to their superiority in access latency and energy consumption, flash memory drives have recently become a competitive alternative to magnetic hard disks as secondary storage. Consequently, flash-based database systems have been receiving increasing attention from the research community in the past few years. In this project, we plan to investigate a number of optimization techniques for transaction management in flash-based database systems by exploiting the characteristics of flash memory drives (such as out-of-place updates and page reprogramming).

Materialized views have been important in improving query evaluation performance in database applications. Since materialized views are managed as real data, they are subjected to both queries and updates. While queries on materialized views are straightforward, updates on the views are not. Recently, materialized views have also played a crucial role in many emerging applications, which often pose new challenges or opportunities to view updates. This calls for a new investigation on the view update problem with these new database applications.

In this project, we investigate innovative query processing algorithms for flash memory based storage-centric sensor networks. Of particular interest are distributed data management issues under sensor system workload given the unique read/write/erase characteristics of flash memories.

This research project proposes to take data-centric designs to improve performance for distributed monitoring systems. Two issues namely filter management for approximate monitoring and routing optimization for in-network data aggregation are studied under this project.

With a location-aware wireless device, a mobile user can query his/her surroundings (e.g., finding the nearest gas station or all shopping centers within 5 miles) at any place, anytime. This project aims to support such location-based services while protecting users' location privacy.

Energy efficiency is a critical consideration in the design of a wireless sensor network. This project exploits the trade-off between data quality and communication cost to improve energy efficiency in sensor data collection.