Hashgraph is asynchronous Byzantine Fault Tolerance (aBFT) - the highest degree of security a consensus algorithm can provide. ABFT means that finality of consensus will be reached with 100% probability if a) attackers control less than 1/3 of the voting power over consensus and b) we assume only that messages from an honest node will eventually get through, but make no assumptions about how long it will take to do so. Specifically, the attacker must control less than 1/3 of the stake in a proof-of-stake system, or less than 1/3 of the nodes in a system without proof-of-stake. The attacker can control the entire communication network in the sense that the attacker can delete messages, or delay messages for arbitrary amounts of time, with the only limitation being that if honest node Alice repeatedly tries to send messages to honest node Bob, eventually one will get through. The system is resilient to attacks on both network nodes and the communication network itself, as long as both types of attacks are within the limits above. Finality of consensus can be contrasted with the probabilistic confidence of proof-of-work systems, where there is always a chance (even if small) of a transaction being retroactively rolled back.
Articles in this section
- How much bandwidth overhead does gossip about gossip add to messages?
- Have any non-Carnegie Mellon professors or academics verified hashgraph as asynchronous Byzantine fault tolerant (ABFT)?
- Who generates the timestamp on a transaction?
- What is 'gossip about gossip and 'virtual voting'?
- Why was hashgraph invented?
- Is the hashgraph consensus algorithm patented?
- How can hashgraph deliver consensus without proof-of-work?
- How efficient is the hashgraph consensus algorithm?
- Does the FLP theorem say aBFT consensus is impossible?
- Where can I find more information about the hashgraph Coq proof?