Asynchronous Byzantine fault tolerance (aBFT) describes a consensus protocol that keeps working correctly even when some participants are malicious and the network makes no promise about how long messages take to arrive. It is the strongest form of Byzantine fault tolerance: many BFT systems assume messages are delivered within some bound, and can stall if that assumption fails, whereas an aBFT protocol only assumes messages are eventually delivered. Like other BFT designs, it tolerates fewer than one third of participants being faulty.
A classic result in distributed computing shows that a fully deterministic protocol cannot guarantee agreement in an asynchronous network, so aBFT protocols rely on randomness to make progress with certainty over time. In crypto, the hashgraph consensus used by Hedera and the Lachesis protocol behind Fantom are often cited as aBFT designs, giving fast and final transaction finality. An early practical example is HoneyBadger BFT (HBBFT), published in 2016, which was designed to keep reaching agreement without any assumptions about network timing.
Ref 1070001097