A Lower Bound for Byzantine Agreement and Consensus for Adaptive Adversaries using VDFs
Large scale cryptocurrencies require the participation of millions of participants and support economic activity of billions of dollars, which has led to new lines of work in binary Byzantine Agreement (BBA) and consensus. The new work aims to achieve communication-efficiency---given such a large $n$, not everyone can speak during the protocol. Several protocols have achieved consensus with communication-efficiency, even under an adaptive adversary, but they require additional strong assumptions---proof-of-work, memory-erasure, etc. All of these protocols use multicast: every honest replica multicasts messages to all other replicas. Under this model, we provide a new communication-efficient consensus protocol using Verifiable Delay Functions (VDFs) that is secure against adaptive adversaries and does not require the same strong assumptions present in other protocols. A natural question is whether we can extend the synchronous protocols to the partially synchronous setting---in this work, we show that using multicast, we cannot. Furthermore, we cannot achieve always safe communication-efficient protocols (that maintain safety with probability 1) even in the synchronous setting against a static adversary when honest replicas only choose to multicast its messages. Considering these impossibility results, we describe a new communication-efficient BBA protocol in a modified partially synchronous network model which is secure against adaptive adversaries with high probability.