What is a 51% attack? How blockchains can be attacked, and what an attacker can't do

Blockchains are often described as unchangeable. That’s mostly true, but it rests on an assumption: that no single party controls most of the network. When that assumption breaks, the result is called a 51% attack. Understanding it explains why exchanges make you wait for confirmations, why small networks are riskier and what the limits of an attack are.

Why a majority matters

Blockchains need a way for many independent computers to agree on one history of transactions without a central authority.

• On proof-of-work networks like bitcoin, miners compete with computing power, and the chain with the most work wins. The bitcoin white paper puts the security assumption plainly: as long as a majority of computing power is controlled by nodes that aren’t cooperating to attack the network, they’ll generate the longest chain and outpace attackers.
• On proof-of-stake networks like Ethereum, validators lock up coins as a stake, and influence depends on how much is staked.

NIST’s blockchain overview lists the risk for both models: the potential for a 51% attack by obtaining enough computational power (proof of work) or enough financial power (proof of stake).

What an attacker could do

With majority control, an attacker can produce blocks faster than everyone else and choose which history the network follows. That allows:

• Double spending: paying someone, getting the goods, then publishing an alternative history in which the payment never happened, so the same coins can be spent again. NIST defines a double spend attack as an attempt to explicitly double spend a digital asset.
• Censorship: refusing to include some people’s transactions.
• Disruption: reorganizing recent blocks, which undermines confidence in the network.

On Ethereum’s proof of stake, ethereum.org sets out the thresholds: an attacker with more than 33% of staked ether could prevent the chain from finalizing; with 51%, they could delay finality, censor and control the chain’s future; with 66% or more, they could finalize their preferred chain and spend, rewind and spend again.

A double spend, step by step

The bitcoin white paper describes the attack it set out to prevent. In simplified form:

1. The attacker pays a merchant, for example sending bitcoin to an exchange.
2. In secret, the attacker builds an alternative chain of blocks in which that payment never happened, sending the same coins back to themselves instead.
3. The merchant waits for some confirmations and releases the goods, or the exchange lets the attacker withdraw other assets.
4. If the attacker’s secret chain becomes longer than the public one, they publish it. Nodes follow the chain with the most work, so the original payment disappears from the accepted history.

Step 4 is only realistic for someone who controls more computing power than the honest network, or gets very lucky over a short period. That’s why the size of the network, and the number of confirmations, matter so much.

What an attacker can’t do

This is where the white paper is reassuring. Even a successful attacker can’t create value out of thin air or take money that never belonged to them: nodes won’t accept invalid transactions as payment, and honest nodes won’t accept blocks containing them. An attacker can only try to change one of their own transactions to take back money they recently spent.

In other words, a 51% attack is mainly a threat to people receiving large payments, not to coins sitting in your wallet.

Why confirmations protect you

Every new block built on top of a transaction is a confirmation. The white paper shows that, as long as honest nodes have the majority, an attacker’s chance of catching up drops exponentially with each block they have to overtake.

That’s why recipients wait. Bitcoin.org describes one confirmation as somewhat reliable, three as mostly reliable and six as the minimum for high-value transfers. Exchanges apply their own confirmation requirements before crediting deposits. See how to send crypto safely.

The numbers from the white paper

The white paper calculates how likely an attacker is to catch up from behind. With an attacker controlling 10% of the computing power, the chance of catching up falls quickly as blocks are added:

It also shows how many confirmations are needed to bring that chance below 0.1%, depending on the attacker’s share:

The pattern is the point: the stronger the attacker, the longer you’d need to wait, and once an attacker has a majority, waiting no longer protects you. These figures are the white paper’s model, not a guarantee for any particular network today.

Why small networks are more at risk

The cost of an attack depends on how much power or stake you need to outweigh everyone else:

• On a large proof-of-work network, gathering most of the computing power would require enormous amounts of hardware and electricity.
• On a small network, far less hardware may be enough to dominate.
• On proof of stake, NIST notes that obtaining the majority of digital assets to “control” the system is generally cost prohibitive, and ethereum.org adds that an attacker’s stake would be put at risk: the community is likely to adopt an honest fork, devaluing the attacker’s coins dramatically.

How proof of stake defends itself

ethereum.org describes several layers of defense:

• Slashing: validators that provably break the rules lose part of their stake. See what is crypto staking?
• Inactivity leak: if a minority stops finalization, the voting power of validators who aren’t participating is gradually reduced until finality can resume.
• Social coordination: as a last resort, the community can coordinate around an honest fork, making an attack economically irrational.

What it means for you

• For everyday holders: a 51% attack on a major network wouldn’t let an attacker take coins from your wallet.
• When receiving large payments: wait for enough confirmations, especially on smaller networks.
• When investing in small coins: remember that weaker security is one more risk. See memecoins explained.

For how consensus works in the first place, see what is a blockchain?, bitcoin mining explained and bitcoin vs Ethereum.