A Novel Scheme for Cache-Aided Multiuser Private Information Retrieval with User-to-User Privacy

Cache-aided Multiuser Private Information Retrieval (MuPIR) is an approach to achieve efficient file retrieval while ensuring multiuser demand privacy against curious servers in multiuser cache-aided PIR systems. More specifically, the demands of the users should be protected from any individual server (i.e., server privacy) during the two-phase retrieval procedure which consists of a cache placement phase and a private delivery phase. One limitation of the MuPIR model is that the users' demands are exposed to each other which is highly undesirable in modern-day distributed retrieval systems where user-to-user privacy has also become an important consideration. Motivated by this, we propose cache-aided MuPIR with user privacy (MuPIR-U), a new problem formulation that simultaneously enforces server privacy and user-to-user privacy. Besides server privacy, it is also required that each user should be prevented from learning other users' demands. We present an optimal scheme for the case of two files and an arbitrary number of users and servers which leverages private cache to achieve demand privacy among users. Interestingly, it is shown that the inclusion of user-to-user privacy incurs no extra download cost in the large memory regime when comparing to MuPIR.