Achieving Similarity Join Services for Outsourced Large Encrypted Datasets

Description

Today, at the rise of big data, a natural technical trend is to outsource the data to public clouds for its cost effectiveness and superior scalability. However, increasing data breaches heighten the concerns on threats of disclosing individual's privacy, since many kinds of data are sensitive, e.g., financial data, genomic data, multimedia data, etc. Although encryption ensures data confidentiality, it prevents clouds from providing useful computations on the data. In this project, we aim to develop a privacy-assured similarity join service over large-scale encrypted datasets, which can be widely used in financial services, bioinformatics, image processing, etc. The proposed service will enable clouds to find pairwise encrypted similar data records without learning the content of the query dataset and the target dataset. The core algorithms will be optimized based on the well-known fast and effective algorithm for similarity search, the latest designs on hash-based data structure, and a practical cryptographic technique called searchable encryption. Our service will contain three deployable software modules for clouds, data owners, and users respectively. This project has great potential for commercialization due to the acute privacy considerations in cloud computing. In addition, it will drive the advancement of other outsourced data-driven services, and also benefit research institutes and common individuals.