Verifiable multi-server private information retrieval

Abstract

Private information retrieval (PIR) allows a client to retrieve any block x i from a database x = x 1 ⋯ x n (stored on a server) such that i remains hidden from the server. PIR schemes with unconditional privacy and sublinear (in n) communication complexity can be constructed assuming multiple honest-but-curious servers. This assumption however cannot be guaranteed in many real life scenarios such as using cloud servers. There are also extra properties such as efficient update of the database. In this paper, we consider a verifiable multi-server PIR (VPIR) model where the servers may be malicious and provide fraudulent answers. We construct an unconditionally t-private and computationally secure k-server VPIR scheme with communication complexity comparable to the best t-private k-server PIR scheme in the honest-but-curious server model. Our scheme supports efficient update of the database, identification of the cheating servers, tolerance of slightly corrupted answers, and multiple database outsourcing. © 2014 Springer International Publishing.