Consensus mechanisms: how blockchains agree
Thousands of strangers, some of them dishonest, agreeing on one list of transactions. Here's how.
A consensus mechanism is the set of rules that lets thousands of independent computers agree on the same order of transactions, even when some of them are faulty or dishonest. Bitcoin uses proof of work, Ethereum uses proof of stake, and many other networks use Byzantine fault tolerant (BFT) voting. They differ mainly in what an attacker must control and in how quickly a transaction becomes final.
The problem consensus solves
Ana has 1 coin and sends it to Ben and, at the same moment, to Carla. Different nodes hear about the two payments in different orders. Without a way to agree which came first, both could be accepted and Ana would have spent the same coin twice. Consensus is how the network picks one history that everyone follows.
Computer scientists call this the Byzantine Generals Problem (Lamport, Shostak and Pease, 1982): generals must agree on a plan by messenger, knowing some generals may be traitors sending conflicting messages. A system that still reaches agreement is "Byzantine fault tolerant".
The main families
- Proof of work (Nakamoto consensus). Block producers spend real computing power; nodes follow the valid chain with the most accumulated work. Anyone can join, and an attacker needs more than half the hashing power. Used by Bitcoin, Litecoin and Dogecoin.
- Proof of stake. Block producers lock up coins as a deposit that can be destroyed (slashed) if they cheat. Ethereum's version, called Gasper, combines a rule for choosing the head of the chain with checkpoint voting for finality.
- Classical BFT voting. A known set of validators votes in rounds, and a block is final once more than two-thirds agree. Practical BFT (Castro and Liskov, 1999) made this usable; Tendermint, now CometBFT, brought it to blockchains such as those in the Cosmos ecosystem. With n validators it tolerates up to one-third being faulty.
| Network | Consensus approach | What an attacker needs |
|---|---|---|
| Bitcoin | Proof of work | More than half the hashing power |
| Ethereum | Proof of stake (Gasper) | A third of staked ETH to stall finality; two-thirds to finalize a bad history |
| Cardano | Proof of stake (Ouroboros) | A majority of stake |
| Solana | Proof of stake, with "Proof of History" as a shared clock | A third of stake to halt it; two-thirds to finalize a bad history |
| Tron | Delegated proof of stake: 27 elected "Super Representatives" | Control of most of the elected block producers |
| XRP Ledger | Federated voting among validators each node chooses to trust | Influence over the trusted validator lists |
| Hedera | Hashgraph (asynchronous BFT), nodes run by a governing council | More than a third of the stake weight behind its nodes |
Finality: when is a payment really done?
Probabilistic finality (Bitcoin). A block is never declared final. Instead, each new block on top makes reversing it exponentially harder. Six confirmations, about an hour, is a common rule of thumb for large amounts; exchanges set their own numbers.
Economic finality (Ethereum). Time is divided into 12-second slots and 32-slot epochs (6.4 minutes). Validators vote on the first block of each epoch. Once a checkpoint has two-thirds of staked ETH behind it twice in a row, it's finalized, usually after about two epochs, roughly 13 minutes. Reverting a finalized block would require validators controlling at least a third of all staked ETH to break the rules provably, and their stake would be destroyed.
Instant finality (classical BFT). Once two-thirds sign a block, it's final immediately. The trade-off: if more than a third go offline, the chain stops rather than risk disagreement.
Safety vs. liveness
Every design chooses what to do when the network splits or many participants vanish. Bitcoin keeps producing blocks (it stays live) and may reorganize later. BFT chains stop until enough validators return (they stay safe). Ethereum does both: blocks keep coming, but finality pauses. That happened twice on May 11 and 12, 2023, when heavy load on some consensus clients stopped finalization for about 25 minutes and then for over an hour; blocks kept being produced and users saw little disruption.
Limitations
- No mechanism stops a majority. Every design assumes most of the power (hashing, stake or votes) is honest. Smaller proof-of-work coins such as Ethereum Classic have suffered 51% attacks.
- Concentration. A few mining pools or staking providers often control large shares, which weakens the guarantees in practice.
- Fast isn't final. A wallet showing "confirmed" after one block isn't the same as finality. For large amounts, wait for the network's finality or several confirmations.
Frequently asked questions
What is the most common consensus mechanism?
By number of networks, proof of stake and its variants. By value secured, proof of work is still huge because Bitcoin uses it.
What is Byzantine fault tolerance?
The ability of a system to reach agreement even when some participants are faulty or lie. Classical BFT systems tolerate up to one-third of participants misbehaving.
How long does Ethereum take to finalize?
Usually about two epochs, roughly 13 minutes. Transactions appear in a block within about 12 seconds, but finality comes later.
How many confirmations should I wait for on Bitcoin?
It depends on the amount. One confirmation is fine for small payments; six (about an hour) is the traditional rule for large ones. Exchanges set their own requirements.
Is proof of stake less secure than proof of work?
They secure the network differently. Proof of work makes attacks cost electricity and hardware; proof of stake makes them cost the attacker's own stake. Both depend on most of the power being honest.
Where to next
Sources
- The Byzantine Generals Problem (Lamport, Shostak and Pease, 1982)
- Practical Byzantine Fault Tolerance (Castro and Liskov, 1999)
- Consensus mechanisms (ethereum.org)
- Gasper (ethereum.org)
- Bitcoin white paper, sections 4–6 and 11
- Ethereum beacon chain patches finality issues (The Block, May 2023)
Written for beginners and checked against the sources listed. Last reviewed November 4, 2026. This page is educational and is not financial, tax or legal advice. How we write and check pages.