International Association
for Cryptologic Research

Anonymous communication is one of the fundamental tools to achieve privacy for communication over the internet. Almost all existing design strategies (e.g., onion routing/Tor, mixnets) for anonymous communication rely on the existence of some honest server/router in the network infrastructure to provide anonymity. A recent seminal work by Shi and Wu (Eurocrypt 2021) proposes the first cryptographic design for a non-interactive anonymous router (NIAR) that can use a single untrusted server or router to permute a set of messages without revealing the permutation to the untrusted router. This work is a really important step towards showing the possibility of designing such protocol from standard cryptographic assumptions. However, their construction is only of theoretical nature and still leaves many open questions towards realizing such systems in practice: (1) the cryptographic building blocks (multi-client functional encryption, correlated pseudorandom function) used in their design are really difficult to implement in practice. (2) Their setup phase takes the permutation as an input to generate the encryption/decryption keys; which means that the messages from the same sender in different rounds will be at the same position in the output vector, unless the setup phase is run before every round with a new permutation. (3) It is not known how to realize such a setup procedure, that initializes a random permutation obliviously, without any trusted entities in the system.

In this paper, we propose the first (somewhat) practical design, which we call sPAR, that solves the above problems using homomorphic encryption techniques. Our design also relies on a one-time setup phase, however the setup phase does not take any specific permutation as input. Instead, our design generates a fresh permutation for every round based on the random values locally generated by the clients. Already existing practical instantiations of fully homomorphic encryption (FHE) schemes make our design implementable and deployable in practice. Our design presents a new direction for designing anonymous communication systems. Unlike some existing systems like Tor, sPAR does not scale to millions of users, however, we demonstrate with a proof-of-concept implementation that sPAR could easily support around hundred users with a few seconds of latency for each message.