More is Merrier: Relax the Non-Collusion Assumption in Multi-Server PIR
Primary Investigator:
Aniket Kate
Tiantian Gong, Ryan Henry, Alexandros Psomas, Aniket Kate
Abstract
A long line of research on secure computation shows that anything that can be computed, can be computed securely using a set of non-colluding parties. This non-collusion assumption is pervasive across secure multi-party computation (MPC). But it remains highly susceptible to covert, undetectable collusion among computing parties. In this work, we relax this traditional assumption in the context of multi-server 1-private PIR (Private Information Retrieval), a crucial aspect of privacy-preserving computations. Traditionally, it operates under the assumption of no pair-wise collusion. Our work introduces a novel collusion deterrence mechanism designed, analyzed, and implemented specifically for 1-private PIR on a public bulletin board, meticulously considering rational and malicious parties.