Private Cell Retrieval From Data Warehouses

Abstract

Publicly accessible data warehouses are an indispensable resource for data analysis. However, they also pose a significant risk to the privacy of the clients, since a data warehouse operator may follow the client’s queries and infer what the client is interested in. Private information retrieval (PIR) techniques allow the client to retrieve a cell from a data warehouse without revealing to the operator which cell is retrieved and, therefore, protects the privacy of the client’s queries. However, PIR cannot be used to hide online analytical processing (OLAP) operations performed by the client, which may disclose the client’s interest. This paper presents a solution for private cell retrieval from a data warehouse on the basis of the Paillier cryptosystem. By our solution, the client can privately perform OLAP operations on the data warehouse and retrieve one (or more) cell without revealing any information about which cell is selected. In addition, we propose a solution for private block download on the basis of the Paillier cryptosystem. Our private block download allows the client to download an encrypted block from a data warehouse without revealing which block in a cloaking region is downloaded and improves the feasibility of our private cell retrieval. Our solutions ensure both the server’s privacy and the client’s privacy. Our experiments have shown that our solutions are practical.