How to achieve adaptive security for asynchronous BFT?

We consider how to build practical asynchronous BFT protocols with adaptive security. In particular, we build two protocols in both the computational setting (where the adversary is limited to polynomial-time) and the stronger information-theoretic model (where the adversary is unbounded). In the computational model, we provide EPIC using adaptively secure key generation and common coin protocols. In the information-theoretical model, we provide HALE leveraging the classic local coin protocol of Bracha. HALE is more robust than EPIC and does not need distributed key generation. Via a five-continent deployment on Amazon EC2, we show EPIC is slightly slower for small and medium-sized networks than the most efficient asynchronous BFT protocols with static security. As the number of replicas is smaller than 46, EPIC's throughput is stable, achieving peak throughput of 8,000–12,500 tx/sec with a transaction size of 250 bytes. When the network size grows larger, EPIC is not as efficient as asynchronous BFT protocols with static security, with throughput of 4,000–6,300 tx/sec. We also show while HALE is in general less efficient than EPIC, HALE is reasonably fast, achieving 42,000 tx/sec and 3,400 tx/sec for the 4-server setting in the LAN/WAN environments, respectively. Remarkably, HALE outperforms EPIC in LANs when the number of replicas is smaller than 16.