Searching Encrypted Data

Company data are very often outsourced to datacentres in order to lower costs of maintaining hardware. If the outsourced data are to be kept secure from a third party, the connection between the datacentre and the company could be secured by a protocol similar to SSL. This, however, requires that the data is stored at the datacentre in plaintext form, meaning the company has to trust the datacentre and its administrators. Alternatively, the data themselves could be encrypted, however, the outputs of typical cryptographic algorithms are not amenable to search. This project explores the research area of searching over encrypted data, specifically using a secure pre-processed index approach. The output of the project is a system developed for the PostgreSQL 8.3 database management system that allows data to be encrypted and stored in a database table whilst allowing for secure, server-side searches with minimal performance overhead, both in space and time. The datarate achieved by the encryption algorithm is around 2.21MB/s, which is shown to be faster than an existing plaintext PostgreSQL indexing algorithm. The search algorithm is shown to be around twice as fast as plaintext linear search, with a rate of 300MB/s, but slower than indexed plaintext search. These results demonstrate the system’s practicality within an industrial setting.