Secure anonymous database search

There exist many large collections of private data that must be protected on behalf of the entities that hold them or the clients they serve. However, there are also often many legitimate reasons for sharing that data in a controlled manner. How can two parties decide to share data without prior knowledge of what data they have? For example, two intelligence agencies might be willing to cooperate by sharing documents about a specific case, and need a way of determining which documents might be of interest to each other. We introduce and address the problem of allowing such entities to search each other's data securely and anonymously. We aim to protect the content of the queries, as well as the content of documents unrelated to those queries, while concealing the identity of the participants. Although there exist systems for solving similar problems, to our knowledge we are the first to address this specific need and also the first to present a secure anonymous search system that is practical for real-time querying. In order to achieve this in an efficient manner, we make use of Bloom filters [5], definitions of security for deterministic encryption [22] that we adapt and instantiate in the private key setting and of a novel encryption primitive, reroutable encryption.