The objective is to design provably secure and scalable schemes for encrypting large-scale databases without losing the ability to query them. Supported by The Modular Approach to Cloud Security (MACS).
Graph databases: I investigate graph encryption. Ideally, a graph encryption scheme should encrypt a graph with support for various graph queries. In this paper, we proposed GRECS that can support approximate shortest distance queries on encrypted graph. See the blog entry. (Collaboration with Seny Kamara)
Relational databases: I also investigate encryption schemes that can provide more functionalities on encrypted relational databases. The challenge is to balance the three objectives of privacy / security, functionality / utility / accuracy, and performance / scalability. For example, in this we propose an efficient secure top-k queries processing on encrypted database.
Multiparty computation: Motivated by the data privacy concerns in today's internet, I am exploring scalable privacy-preserving data mining/machine learning protocols where each party only learns the output of the algorithm and nothing else. In particular, I consider mostly data mining/learning algorithms that can scale to massive amount of datasets. For example, I proposed a privacy-preserving protocol for computing the distance between two time-series (see here). I am also studying privacy-preserving clustering algorithms. (Collaboration with RSA labs)
Node similarity: I am interested in node similarity measures, a fundamental problem in graph data analytics that can be used for graph de-anonymization. Existing node similarity measurements focus on evaluating how similar between two nodes in the same graph. In this project, we investigate methods that measure similarity between nodes in different graphs (inter-graph nodes). In particular, we are interested in defining new distance metrics that are based on the neighborhood topology of two nodes.