A 51% attack is majority control of proof-of-work block production that can enable recent reorganizations, double-spending, or transaction censorship.
A 51% attack is a proof-of-work consensus attack in which one actor or coordinated group controls enough effective hash power to outpace the honest network and dominate the chain selected by nodes. The name refers to majority control, but attack feasibility is not a perfectly fixed threshold: an attacker below 50% can sometimes cause disruption, while sustained majority hash power makes repeated reorganizations and censorship substantially more practical.
The attacker does not gain every power over the blockchain. Validating nodes still apply protocol rules, and private keys remain necessary to spend other users’ assets. The principal risk is control over the ordering and recent history of otherwise valid transactions.
| Capability | Why majority hash power helps | Important limit |
|---|---|---|
| Reorganize recent blocks | A privately mined branch can accumulate more proof of work than the public branch | Cost and difficulty rise as the targeted history becomes deeper |
| Double-spend the attacker’s funds | A payment on the displaced branch can be replaced by a conflicting transaction on the winning branch | The attacker needs valid authority over the spent funds |
| Censor selected transactions | Dominant block production can omit targeted transactions | Other miners may include them when dominance ends |
| Delay confirmations | The attacker can disrupt normal block ordering and inclusion | It cannot make an invalid transaction valid |
| Capture block rewards | Dominant production can win more blocks while the attack persists | Rewards may not offset energy, equipment, rental, detection, and price-impact costs |
The distinction matters: proof of work helps nodes choose among valid competing histories. It does not replace transaction-signature and validation rules.
flowchart TD
A["Attacker sends a valid payment to a merchant"] --> B["Merchant observes confirmations on the public chain"]
B --> C["Attacker privately mines a competing valid branch"]
C --> D["Competing branch excludes or replaces the attacker's payment"]
D --> E["Attacker publishes a branch with more accumulated work"]
E --> F["Nodes reorganize to the selected valid branch"]
F --> G["Merchant's confirmed payment is no longer in canonical history"]
This does not duplicate arbitrary third-party funds. It lets the attacker attempt to keep both the goods received from the merchant and the attacker’s own funds by replacing the history containing the original payment.
A merchant accepts a digital asset worth $20,000 after two block confirmations and releases an irreversible off-chain asset. The payment network has low total hash power, and comparable mining capacity can be rented temporarily.
While the merchant waits, the payer privately mines a competing branch containing a conflicting transaction that returns the payer’s funds to another address the payer controls. The attacker publishes that branch after it accumulates more proof of work than the branch the merchant observed. If validating nodes select the competing branch, the merchant’s payment disappears from canonical history even though it had appeared confirmed.
The merchant loses the off-chain asset and no longer has the expected on-chain payment. Waiting for more confirmations would require the attacker to replace more accumulated work and can reduce the probability of success, but it cannot guarantee finality under every concentration and cost scenario.
Technical capability is only part of the analysis. An attacker may need to fund:
Potential targets include exchanges, bridges, merchants, and custodians that release another asset after too few confirmations. The economically relevant exposure can exceed the value transferred on-chain if the victim delivers cash, securities, inventory, or another token that cannot be reversed.
No single metric proves safety. A high reported hash rate can still be concentrated, while a broad set of named pools may depend on fewer underlying operators or infrastructure providers.
Proof-of-stake networks can face majority-control, censorship, reorganization, governance, or finality attacks, but hash power is not the controlling resource. Stake thresholds, slashing, validator rules, checkpoints, and social recovery differ by protocol. Calling every consensus attack a 51% attack can hide those distinctions.
A credible attack can lead exchanges to suspend deposits and withdrawals, increase confirmation requirements, or delist an asset. Market makers may widen spreads, merchants may stop accepting payment, and users may demand compensation for greater settlement uncertainty.
These responses can reduce liquidity and market confidence even when no individual wallet key is compromised. Network security, custody security, and asset-price risk are related but separate layers.
This article provides general financial and technical education, not individualized cybersecurity, custody, legal, trading, or investment advice.