Private data indexes for selective access to outsourced data

Cloud storage services have recently emerged as a successful approach for making resources conveniently available to large communities of users. Several techniques have been investigated for enabling such services, including encryption for ensuring data protection, as well as indexing for enabling efficient query execution on encrypted data. When data are to be made available selectively, the combined use of the two techniques must be handled with care, since indexes can put the confidentiality protection guaranteed by encryption at risk. In this paper, we investigate this issue and propose an indexing technique for supporting efficient access to encrypted data while preventing possible disclosure of data to users not authorized to access them. Intuitively, our indexing technique accounts for authorizations when producing indexes so to ensure that different occurrences of the same plaintext value, but accessible by different sets of users, be not recognizable from their indexes. We show that our solution exhibits a limited performance overhead in query evaluation, while preventing leakage of information.