It Takes Two to Infer: Partitioning as an Encrypted Database Defense

As more organizations and companies move to the cloud, the need to protect against data breaches becomes ever more important. Although previous systems using encrypted databases have been developed, the intricacies of their design lead to further attacks that compromise the privacy of the stored data. We propose PartDB, a system to protect against such attacks. PartDB partitions data in a reconstructable manner to prevent an attacker from recovering individual rows of a database, while still allowing an application to treat the system as a single whole. We investigate multiple threat models, involving various degrees of compromise of the database system, and provide defenses countering each threat.