Blinder: MPC Based Scalable and Robust Anonymous Committed Broadcast

Anonymous Committed Broadcast is a functionality that extends DC-nets and allows a set of clients to privately commit a message to set of servers, which can then simultaneously open all committed messages in a random ordering. Anonymity holds since no one can learn the ordering or the content of the client’s committed message. We present Blinder, the first system that provides a scalable and fully robust solution for anonymous committed broadcast. Blinder maintains both properties of security (anonymity) and robustness (aka. ‘guaranteed output delivery’ or ‘availability’) in the face of a global active (malicious) adversary. Moreover, Blinder is censorship resistant, meaning that a honest client cannot be blocked from participating. Blinder obtains its security and scalability by carefully combining classical and state-of-the-art techniques from the fields of anonymous communication and secure multiparty computation (MPC). Relying on MPC for such a system is beneficial since it naturally allows the parties (servers) to enforce some properties on accepted messages prior their publication. In order to demonstrate scalability, we evaluate Blinder with up to 1 million clients, up to 100 servers and a message size of up to 10 kilobytes. In addition, we show that it is a perfect fit to be implemented on a GPU. A GPU based implementation of Blinder with 5 servers, which accepts 1 million clients, incurs a latency of less than 8 minutes; faster by a factor of > 100 than the 3-servers Riposte protocol (SOSP ’15), which is not robust and not censorship resistant; we get an even larger factor when comparing to AsynchroMix and PowerMix (CCS ’19), which are the only constructions that guarantee fairness (or robustness in the online phase).