Glossary · Consensus

Proof of Authority (PoA)

Published 23 May 2025 · Updated 2 Nov 2026 · 15 min read

Proof of Authority (PoA) is a consensus method where approved accounts, called authorities, validators or signers, take turns producing and signing blocks. This fixed list sits in the genesis block or a contract. Members' votes, a governing body or a stake election can change the list. Participants risk identity, reputation or membership rather than mining power or stake.

  • PoA's allowlist cuts mining and staking costs, but stolen keys breached Ronin's 5-of-9 bridge in March 2022.
  • Clique needs a simple majority online, while QBFT needs at least two thirds to finalize blocks.
  • In November 2026, Ethereum's PoA testnets are gone, while QBFT, Aura and BNB Smart Chain's PoSA still run.
How Proof of Authority worksClique, Aura, IBFT 2.0, QBFT and PoSAHow many failures a PoA network survivesProof of Authority versus stake and workWhere PoA runs and where it endedWhat PoA does not protect againstWhen PoA fits and when it does notHow Proof of Authority developedDesigning a permissioned consensus networkProof of Authority FAQ

How Proof of Authority works

Only approved accounts can propose the next block. Other participants check each signature against that list. In this consensus method, the first authorities appear in the genesis block or a contract. Each authorized validator follows the protocol's production rules. Clique rotates signers and lets them vote on membership changes. PoA gets Sybil resistance from this allowlist, not from making new identities expensive.

Figure 1 · Clique rules from EIP-225Explore a signing rotationChoose 3, 5, 7 or 9 signers, then take some offline to see the next 12 blocks.
Signers
0 offline
SignersAOnlineBOnlineCOnlineDOnlineEOnlineFOnlineGOnline
Who signs the next 12 blocks#2,520AIn turn#2,521BIn turn#2,522CIn turn#2,523DIn turn#2,524EIn turn#2,525FIn turn#2,526GIn turn#2,527AIn turn#2,528BIn turn#2,529CIn turn#2,530DIn turn#2,531EIn turnIn turnOut of turnNo signer allowed
Signers online7 of 7Each may sign 1 block in every 4
Needed to keep going4Half of 7 rounded down, plus one
Out-of-turn blocks0 of 12They score difficulty 1 instead of 2

Every signer is online, so the strip follows the scheduled in-turn rotation across all 12 blocks.

Some signers are offline, but enough remain eligible to keep producing blocks out of turn.

Too few signers remain online to satisfy the recent-signing rule, so production stops.

Rules follow EIP-225 and go-ethereum's Clique code. The strip simplifies which out-of-turn signer wins.

For N signers, block B assigns its in-turn producer at index B modulo N. In-turn blocks score difficulty 2, while out-of-turn blocks score 1. Each signer may sign at most once per floor(N/2)+1 consecutive blocks. Adding or removing a signer needs floor(N/2)+1 votes, which reset at each epoch checkpoint, every 30,000 blocks by default. Progress requires floor(N/2)+1 online, so 7 signers survive 3 offline and 4 survive only 1. The highest-difficulty chain wins, without immediate finality.

Aura assigns one producer to each fixed time step. For step s and n authorities, the primary has index s modulo n. Only that primary may issue a block. A block becomes final after more than half the authorities have built on it. Energy Web Chain shows this rotation in practice. A 50-block sample from 2 November 2026 shows six authorities sealing blocks every 5 seconds. One missed a step, and the next continued.

Clique, Aura, IBFT 2.0, QBFT and PoSA

A permissioned validator set does not specify a single algorithm. Proof of Authority includes rotating producers and protocols that vote before accepting a block. GoQuorum documentation also lists crash-tolerant Raft under PoA.

ProtocolHow blocks are madeFinalityStatus in 2026
CliqueHow blocks are madeSigners take turns by block number, while others may sign out of turnFinalityAccepted blocks can be replaced through chain reorganizationsStatus in 2026go-ethereum stopped signing in April 2024. Besu removed it in April 2026
AuraHow blocks are madeOnly the primary assigned to a time step may produce a blockFinalityBlocks become final after a majority of authorities build on themStatus in 2026Energy Web Chain runs this protocol in November 2026
IBFT 2.0How blocks are madeA block needs signatures from at least two thirds of validatorsFinalityEach accepted block has immediate finality under the fault assumptionsStatus in 2026Besu supports it for existing private networks in November 2026
QBFTHow blocks are madeBesu requires a quorum equal to the ceiling of 2n/3FinalityConflicting blocks cannot finalize within the protocol's Byzantine fault limitStatus in 2026Besu recommends this protocol for private networks in November 2026
Proof of Staked AuthorityHow blocks are madeEach epoch, 21 of the 45 stake-elected validators take turns, with 0.45-second blocksFinalityBNB Smart Chain reports about 1.125 seconds with fast finality in November 2026Status in 2026BNB Smart Chain, with a daily stake election

How many failures a PoA network survives

Voting protocols wait for a quorum before accepting a block. In Besu, that means signatures from at least two thirds of the validator set, rounded up. A crashed validator stops responding. A malicious validator may sign conflicting blocks. Byzantine fault tolerance accounts for malicious behavior. Within its fault limits, a permissioned BFT PoA protocol commits each block without waiting for later confirmations.

Figure 2 · QBFT and IBFT 2.0 quorum rulesTest quorum and failure limitsSet 1 to 25 validators and adjust crashed and malicious counts to compare QBFT thresholds with Clique.
7 validators
0 crashed
0 malicious
ValidatorsBlocks finalizeHonest and onlineMaliciousCrashedEach block gets 5 signatures from honest validators and is final at once.
Quorum5 of 7Signatures each block needs, two thirds rounded up
Crashes tolerated2Offline validators before blocks stop
Malicious tolerated2With n at least 3f + 1, as in classic BFT

Clique with 7 signers keeps going with 4 online, so it survives 3 offline. Its blocks are never final at once and can be replaced in a fork.

Enough honest validators remain online to reach quorum, and malicious participation stays below the conflicting-decision threshold.

Honest validators cannot reach quorum alone. If malicious validators withhold signatures, the chain cannot progress.

Malicious signatures can bridge two quorums, allowing conflicting blocks to finalize if both groups receive enough signatures.

Malicious validators alone reach quorum, so they can approve blocks without any honest validator's signature.

With three or fewer validators, Besu gives no guarantee against adversarial behavior, even if signatures reach quorum.

The quorum is the ceiling of 2n/3 from Hyperledger Besu's source. The Clique row follows EIP-225.

With seven validators, Clique continues with three offline, while QBFT needs five signatures. The former can suffer a chain fork. The latter stops without quorum and prevents conflicting decisions within its fault limits. For QBFT with 21 validators, seven crashes are tolerable, but only six malicious validators are. Besu requires at least four validators for Byzantine fault tolerance. Larger QBFT and IBFT 2.0 groups need more signatures per block, and their block time grows.

Proof of Authority versus stake and work

The main difference is what gives a producer permission to participate. Proof of Authority uses membership rules. Proof of Stake uses deposited assets, while Proof of Work uses computational effort. Low energy use is not unique to PoA.

AspectProof of AuthorityProof of StakeProof of Work
Who produces blocksProof of AuthorityApproved accounts produce blocks according to their network's rulesProof of StakeValidators who deposit at least 32 ETH and run client softwareProof of WorkBitcoin miners compete to find a hash below the target
What a producer risksProof of AuthorityAuthorities risk their reputation, identity or continued network membershipProof of StakeEthereum penalizes conflicting signatures by slashing the deposited stakeProof of WorkBitcoin miners spend electricity and hardware resources while competing
How new producers joinProof of AuthorityAdmission requires a membership vote, operator-controlled contract or PoSA electionProof of StakeEthereum admits participants through a deposit and an activation queueProof of WorkAnyone with suitable hardware can join Bitcoin mining
FinalityProof of AuthorityBFT PoA protocols commit blocks immediately within their fault limitsProof of StakeEthereum takes about 15 minutes to finalize a block in November 2026Proof of WorkEach additional Bitcoin block makes a reversal less likely
Cost and energyProof of AuthorityPoA without staking elections requires neither mining nor purchased stakeProof of StakeA 2022 estimate places Ethereum's annual electricity use near 0.0026 TWhProof of WorkA 2025 Cambridge estimate places Bitcoin mining's annual electricity use at 138 TWh

Where PoA runs and where it ended

Early deployments served permissioned chains and Ethereum testing. A blockchain testnet needs a way to authorize production even when its assets have no value. Proof of Authority provided that control through approved keys. Several public networks later replaced PoA with staking. Kovan, Rinkeby and Görli are gone in November 2026, while Energy Web Chain and BNB Smart Chain still run authority-based consensus.

Figure 3 · Eight networks, 2017 to 2026Follow each network's PoA yearsSelect a network to see its PoA timeline, production rules, operator selection and any later consensus changes.
Runs PoA or PoSAMoved to stakingShut down
Kovan testnetShut down
Engine
Aura (Parity)
PoA years
March 2017 to June 2022
Block time
4 seconds
Authorities
9 at launch, later 3 and a contract
Chosen by
A consortium of 10 Ethereum companies
What happened
Deprecated by June 2022 after OpenEthereum development stopped
Rinkeby testnetShut down
Engine
Clique (EIP-225)
PoA years
April 2017 to mid 2023
Block time
15 seconds
Authorities
3 signers at genesis
Chosen by
Votes of existing signers in block headers
What happened
Not moved to proof of stake, removed from go-ethereum in June 2023
VeChainThorMoved to staking
Engine
PoA, then PoA 2.0
PoA years
June 2018 to December 2025
Block time
10 seconds
Authorities
Up to 101 Authority Masternodes
Chosen by
The VeChain Foundation steering committee, after KYC
What happened
Hayabusa began the move to delegated proof of stake on 2 December 2025
xDai and Gnosis ChainMoved to staking
Engine
Aura, with POSDAO from December 2020
PoA years
October 2018 to December 2022
Block time
5 seconds
Authorities
Up to 19 validators
Chosen by
POSDAO, with a minimum of 2,000 GNO staked
What happened
Merged with its beacon chain on 8 December 2022
Görli testnetShut down
Engine
Clique (EIP-225)
PoA years
January 2019 to August 2022
Block time
15 seconds
Authorities
1 signer at genesis
Chosen by
Votes of existing signers in block headers
What happened
Moved to proof of stake in August 2022, sunset on 13 April 2024
Energy Web ChainLive PoA
Engine
Aura
PoA years
June 2019 to today
Block time
5 seconds
Authorities
10 at launch, 6 sealing blocks on 2 November 2026
Chosen by
Legally registered members of the Energy Web Foundation
What happened
Still runs Proof of Authority in November 2026
BNB Smart ChainLive PoSA
Engine
Parlia, Proof of Staked Authority
PoA years
September 2020 to today
Block time
3 seconds at launch, 0.45 seconds since January 2026
Authorities
21 at launch, 45 now (21 Cabinets and 24 Candidates)
Chosen by
A daily election by stake
What happened
Validators halted the chain after the October 2022 bridge exploit
RoninMoved to staking
Engine
PoA, then delegated proof of stake
PoA years
May 2021 to April 2023
Block time
3 seconds
Authorities
9 validators in March 2022
Chosen by
Sky Mavis and partner organizations
What happened
5 of 9 keys were stolen in March 2022. An OP Stack layer 2 since May 2026

Kovan launched with Aura in March 2017, used 4-second steps, and was already deprecated in June 2022.

Rinkeby began in April 2017 with three unnamed signers and 15-second blocks. go-ethereum removed it in June 2023.

VeChainThor used up to 101 Authority Masternodes. Its 2 December 2025 Hayabusa upgrade began the move to delegated proof of stake.

xDai launched in October 2018 with 5-second blocks. Gnosis Chain left Aura through its 8 December 2022 merge.

Görli began with one approved signer in January 2019, switched to staking in August 2022, and reached its sunset date on 13 April 2024.

Energy Web Chain uses Aura in November 2026. Its operators must be legally registered organizations that are members of the Energy Web Foundation.

BNB Smart Chain launched PoSA in September 2020. In November 2026, daily stake elections choose 21 Cabinets and 24 Candidates.

Ronin began with PoA in May 2021, adopted delegated staking in April 2023, and became an OP Stack layer 2 on 12 May 2026.

Dates come from network documentation, client code and announcements, checked on 2 November 2026.

BNB Smart Chain's PoSA has 45 validators and 0.45-second blocks in November 2026. Energy Web Chain runs Aura, while LACChain and Alastria's Red B use IBFT 2.0 on Besu. Clique support has shrunk: go-ethereum stopped signing its blocks in April 2024, and Besu removed the protocol in April 2026. GoQuorum's repository was archived on 5 June 2026, and Consensys no longer supports the client.

What PoA does not protect against

PoA decides who may sign blocks, not how keys are stored. The Ronin and Harmony incidents concern bridge signing thresholds, not the chain's block production. An approved operator can also censor transactions, and a permitted key can be copied or stolen.

Stolen authority keys

Ronin's bridge threshold was five of nine in March 2022. An attacker gained four Sky Mavis keys through spear phishing and one Axie DAO key through an allowlist not revoked after December 2021.

Low signing thresholds

In the 23 June 2022 Harmony Horizon theft, attackers used compromised private keys. Reports described a two-of-five threshold, which Harmony raised to four of five.

Halts and blocklists

After the 6 October 2022 BSC Token Hub exploit, BNB Smart Chain validators suspended production. A client update then blocked the attacker's addresses at the consensus level.

Censorship by rule

VeChainThor introduced a consensus-level blocklist through a fork on 10 January 2020. In November 2026, its client still lists 469 blocked addresses.

Cloned signing keys

A 2020 NDSS study demonstrated a clone attack against Aura and Clique. One authority copied its key across nodes communicating with different groups, splitting the network.

Shrinking client support

Version 1.14.0 of go-ethereum dropped Clique signing on 24 April 2024. Besu 26.4.0 removed it in April 2026. GoQuorum is unsupported in November 2026.

When PoA fits and when it does not

PoA suits groups that accept a named operator set for a permissioned ledger or private blockchain. The chosen protocol, such as Clique or QBFT, determines how many failures the network survives.

Where PoA fits

  • Shared consortium controlKnown companies can distribute voting rights, node operation and upgrade approval. BFT PoA can provide immediately committed blocks within its fault limits.
  • Private testing environmentsTeams can test with an approved signing list without relying on valuable mining rewards. The chosen client's protocol support still matters.
  • Named network operatorsWhere legal identity is a membership requirement, PoA can restrict production accordingly. The network must define that requirement explicitly.
  • Trusted operator sidechainsA sidechain can fit this model when participants accept its approved operator group. Restricted production remains part of the trust agreement.

Where it does not

  • Open producer admissionA public network that requires unrestricted entry cannot provide it through a closed PoA allowlist. Admission remains a governance decision.
  • Concentrated key controlSeveral keys do not necessarily represent independent operators. A design with few operators or concentrated custody must account for correlated compromise.
  • Neutrality toward usersApproved producers can censor transactions. A project requiring neutrality cannot treat restricted membership as proof that every valid transaction will be included.
  • Production maintenance concernsIn November 2026, reduced client support makes Clique a maintenance concern for new production networks. Protocol selection must include support availability.

How Proof of Authority developed

The term dates to private chains described in November 2015. Ethereum's test networks later made PoA widely known. Client support is narrower in November 2026, while some public networks still run authority-based consensus.

  1. November 2015

    Private chains get a guide

    Gavin Wood's TurboEthereum guide described secret keys collaborating to create the longest chain. The text used the term Proof of Authority for private chains.

  2. March 2017

    A testnet and protocol emerge

    A consortium of 10 Ethereum companies launched Kovan on 6 March 2017. EIP-225 introduced Clique the same day, following Ropsten's February 2017 attacks.

  3. April 2017

    The first EIP-225 network launches

    Rinkeby's genesis dates to 12 April 2017. It launched with three unnamed signing keys and a 15-second block interval.

  4. December 2017

    POA Network requires notary licenses

    POA Network launched on 15 December 2017 with 12 validators. Each validator held a US notary public license.

  5. June 2019

    Energy Web Chain launches

    Energy Web Chain launched with 10 validators. It runs Aura in November 2026, with membership limited to legally registered organizations that are members of the Energy Web Foundation.

  6. September 2020

    BNB Smart Chain launches

    BNB Smart Chain launched on 1 September 2020 with Proof of Staked Authority. Its initial 21 validators produced a block every 3 seconds.

  7. April 2024

    Signing ends in go-ethereum

    The go-ethereum v1.14.0 release on 24 April 2024 removed the client's ability to sign blocks on Clique networks.

  8. April 2026

    Besu 26.4.0 drops support

    Besu 26.4.0 removed Clique in April 2026. GoQuorum's repository was archived on 5 June 2026, and Consensys no longer supports it in November 2026.

Designing a permissioned consensus network

We are a software engineering company founded in 2007, with blockchain work dating to 2013. Our consortium blockchain development covers consensus, membership and data-scoping decisions. Voting rights, node operation and upgrade approval can be distributed so no single member decides for the group. For a permissioned PoA project, those rules help define the trust model rather than leaving it implicit.

PixelPlex case icon

We built a digital contract platform for Kazakhtelecom's contract workflows using Quorum, an Ethereum-based framework supporting Solidity smart contracts. The platform separates private and public data. Private smart contracts hold client data. Quorum uses modified consensus for internal, non-public projects.

  1. Protocol and validator count

    We help define consensus requirements and membership rules. The design must distinguish the protocol's crash tolerance from its tolerance for malicious participation.

  2. Operator roles and topology

    We address node topology through hybrid network planning. The operator or consortium sets membership rules, the validator set and upgrade rights.

  3. Key custody and thresholds

    A design review should separate who controls signing keys from how many signatures are required. Several keys held together do not create independent control.

  4. Membership and upgrade governance

    We help distribute voting rights and upgrade approval across members. Admission and removal rules determine how that shared control changes over time.

  5. Monitoring and incident response

    Monitor missed turns or rounds, changes to the signer or validator set, and client versions. Define responses to stalled production and suspected key compromise.

Discuss the rules for your permissioned network

We can discuss membership, consensus and data boundaries, then identify the decisions your operator group needs to make.

Discuss your network

Proof of Authority FAQ

What is Proof of Authority in simple terms?

Proof of Authority is a way to run a blockchain where only a fixed list of approved accounts may produce blocks. The list sits in the genesis block or a contract, and existing authorities usually vote to change it. Energy Web Chain's six authorities seal a block every 5 seconds in the 2 November 2026 sample.

How does PoA differ from Proof of Stake?

PoA admits block producers by approval, while Proof of Stake requires a deposit that can be slashed. Ethereum requires at least 32 ETH and admits anyone through a deposit and activation queue in November 2026. Proof of Staked Authority combines these models through staking elections for a limited rotating group.

Is a PoA network actually decentralized?

PoA is less decentralized than open Proof of Stake or Proof of Work because a small, fixed group produces all blocks. Becoming an authorized signer is typically out of reach for an ordinary person. BNB Smart Chain has 45 validators in November 2026.

What are the advantages and disadvantages of PoA?

Ordinary PoA is cheap to run, avoids mining and stake purchases, and supports fast blocks. QBFT and IBFT 2.0 give immediate finality within their fault limits. Disadvantages include few producers, censorship and stolen keys, as Ronin's bridge attack showed in March 2022. Clique also has shrinking client support in November 2026.

Which blockchains use Proof of Authority today?

In November 2026, Energy Web Chain uses Aura, while BNB Smart Chain combines staking elections with authority-style rotation. LACChain and Alastria's Red B use IBFT 2.0 on Besu. Ethereum's former PoA test networks, Kovan, Rinkeby and Görli, are shut down.

What happens if a PoA validator is compromised?

It depends on how many keys are compromised and on the protocol's threshold. A compromised key gives an attacker that member's signing power, not automatically control of the whole network. With nine participants, Clique's majority is five, while a QBFT quorum is six. The Ronin bridge attacker gained five of nine signing keys in March 2022, meeting its withdrawal threshold.

How do PoA, QBFT and IBFT differ?

Proof of Authority is the family, and IBFT 2.0 and QBFT are two BFT protocols in it. Both require signatures from at least two thirds of validators and provide immediate finality. Besu recommends QBFT for private networks and supports IBFT 2.0 for existing ones in November 2026. Clique and Aura are PoA protocols without immediate finality.

Related terms

All definitions
Sources (32)

Get updates about blockchain, technologies and our company

We will process the personal data you provide in accordance with our Privacy policy. You can unsubscribe or change your preferences at any time by clicking the link in any email.

Follow us on social networks and don't miss the latest tech news

  • facebook
  • X
  • linkedin
  • instagram
Stay tuned and add value to your feed