A 51% attack is an attack on a blockchain in which a single party or coordinated group controls more than half of the resource that secures consensus, such as hash rate in proof of work or staked value in proof of stake, and uses that majority to rewrite recent history, reverse its own payments or censor other users. It is the central threat model that consensus mechanisms are designed to make too expensive to attempt.
The name refers to the majority threshold, not a precise number. An attacker with somewhat less than half of the hash rate can still succeed with meaningful probability over short horizons, while an attacker with a large majority can succeed almost with certainty. The danger is greatest on smaller networks, where renting or acquiring a majority costs relatively little.
How a 51% attack works in proof of work
Proof of work networks follow Nakamoto consensus: nodes accept the valid chain with the most accumulated work. An attacker with majority hash rate can therefore mine a private chain in secret, publish it later, and have the network adopt it because it represents more work than the public chain.
- The attacker sends coins to an exchange or merchant on the public chain.
- At the same time, the attacker mines a hidden competing chain in which those coins go back to an address the attacker controls.
- The exchange waits for its required confirmations and credits the deposit, which the attacker then trades and withdraws.
- The attacker releases the hidden chain, which is longer, and nodes reorganize to it.
- The original deposit disappears from history, leaving the exchange with the loss: a double spend.
The attack is economically motivated. It succeeds only if the value of the double spend exceeds the cost of the hash rate plus any loss in value of the attacker’s own holdings and equipment. That calculation is why the same attack that is impractical on a large network can be profitable on a small one.
What a majority attacker can and cannot do
Majority control is powerful but bounded by protocol rules. Every honest node still validates every block, so an attacker cannot create transactions that break the rules, regardless of hash rate or stake. The table separates realistic capabilities from common misconceptions.
| Action | Possible | Explanation |
|---|---|---|
| Double-spend own coins | Yes | Reorganizing history reverses the attacker’s earlier payment |
| Censor transactions | Yes | Attacker can exclude chosen transactions from its blocks |
| Orphan honest blocks | Yes | Competing miners lose rewards on replaced blocks |
| Steal others’ coins | No | Valid signatures from the owner are still required |
| Create coins beyond the rules | No | Nodes reject blocks that violate issuance rules |
| Change old history cheaply | No | Cost grows with the depth of blocks to rewrite |
The defining risk is therefore reversal of recent transactions and censorship, not theft of arbitrary balances. That distinction matters for exchanges and merchants, who are the usual victims because they release goods or funds after a deposit appears final.
Majority attacks under different consensus mechanisms
Every consensus mechanism has a majority threshold, but the resource, the cost and the consequences differ. Byzantine fault tolerant protocols, for example, typically assume that fewer than one third of voting power is faulty, so a party with one third can halt finality and a party with two thirds can finalize conflicting history.
| Mechanism | Resource to control | Critical threshold | Main deterrent |
|---|---|---|---|
| Proof of work | Hash rate | Above 50% for reliable reorganization | Hardware and energy cost |
| Proof of stake | Staked tokens | One third to halt, two thirds to finalize falsely | Slashing of attacker stake |
| Delegated proof of stake | Votes for block producers | Majority of elected producers | Voter replacement of producers |
| Proof of authority | Approved validator identities | Majority of validators | Legal and reputational accountability |
In proof of stake, slashing changes the economics. Validators who sign conflicting blocks can lose a large part of their stake, so an attacker risks destroying the capital used for the attack. Proof of work offers no comparable penalty, although specialized mining hardware loses value if the attacked coin collapses.
Factors that set the cost of an attack
The cost of a 51% attack depends less on the protocol than on the size and structure of its security resource. A network that shares its mining algorithm with a much larger network is especially exposed, because hash rate can be rented or redirected from the larger network for a few hours.
| Factor | Raises attack cost when | Lowers attack cost when |
|---|---|---|
| Total hash rate or stake | It is large and widely distributed | It is small |
| Algorithm sharing | Hardware is unique to the network | A larger chain uses the same algorithm |
| Rental market | Little rentable capacity exists | Hash rate can be rented by the hour |
| Mining pool concentration | Many independent pools | Few pools control most blocks |
| Slashing rules | Attackers lose substantial stake | No penalty for equivocation |
Smaller proof of work coins have suffered real double-spend reorganizations that cost exchanges significant sums. Larger networks are not immune in principle, but the hardware and energy required to outpace them make a direct attack prohibitively expensive under current conditions.
Defenses against 51% attacks
Defense combines protocol design with operational caution by the parties that receive payments. The most effective tool for exchanges is simply to wait for more confirmations on smaller networks, sized to the value at risk and the cost of renting enough hash rate to reverse that many blocks.
- Confirmation depth: require deeper confirmations on chains with low hash rate or high rental availability.
- Finality gadgets: protocols that make blocks irreversible after a vote reduce reorganization risk.
- Reorganization monitoring: alerting on unusual chain reorganizations allows deposits to be frozen quickly.
- Decentralized mining and staking: spreading power across many independent operators raises coordination costs.
The idea in brief
A 51% attack happens when one group gains control of most of the power that keeps a blockchain honest. With that majority, it can quietly build its own version of recent history and then make the network accept it instead of the real one.
Imagine a town where decisions are made by majority vote. Whoever controls most votes can overturn yesterday’s decisions, though not break the town’s basic laws. On a blockchain, that means the attacker can undo its own recent payments, but cannot take coins from other people’s accounts.
A practical scenario
A cryptocurrency exchange lists a small proof of work coin that shares its mining algorithm with a much larger network. Its risk team estimates that renting enough hash rate to dominate the small coin for several hours would cost far less than a large deposit on the exchange.
The team raises the required confirmations for that coin from a handful of blocks to several hundred, sets lower withdrawal limits for new deposits and adds an alert for chain reorganizations deeper than a few blocks. Weeks later the alert fires: a reorganization replaces dozens of blocks. Because the exchange had not yet credited the affected deposits, the reversed transactions cause no loss, and the team pauses deposits until the network stabilizes.