Rabbit-Mix: Robust Algebraic Anonymous Broadcast from Additive Bases
Chongwon Cho, Samuel Dittmer, Yuval Ishai, Steve Lu, Rafail Ostrovsky
Abstract:
We present Rabbit-Mix, a robust algebraic mixing-based anonymous broadcast protocol in the client-server model. Rabbit-Mix is the first practical sender-anonymous broadcast protocol satisfying both robustness and 100% message delivery assuming a (strong) honest majority of servers. It presents roughly 3x improvement in comparison to Blinder (CCS 2020), a previous anonymous broadcast protocol in the same model, in terms of the number of algebraic operations and communication, while at the same time eliminating the non-negligible failure probability of Blinder. To obtain these improvements, we combine the use of Newton's identities for mixing with a novel way of exploiting an algebraic structure in the powers of field elements, based on an additive 2-basis, to compactly encode and decode client messages. We also introduce a simple and efficient distributed protocol to verify the well-formedness of client input encodings, which should consist of shares of multiple arithmetic progressions tied together.
comment:
USENIX Security Symposium 2024: 3151-3168
Fetch PDF file of the paper