The 51% attack

An attacker with most of the hash power could rewrite recent blocks and double spend their own coins, but could not steal coins or change Bitcoin's rules.

The race

If an attacker controls more hash power than the rest of the network combined, they can mine a secret chain faster than everyone else, then release it. Nodes follow the chain with the most work, so the attacker's chain replaces recent blocks.

What an attacker could do

• Double spend their own coins: pay a merchant, wait for confirmations, then publish a chain where that payment never happened.
• Censor transactions by refusing to include them, and orphaning other miners' blocks.

What they couldn't do

• Spend coins they don't have the keys for.
• Create more bitcoin than the rules allow.
• Change any rule: nodes reject invalid blocks regardless of hash power.

Why it's unlikely

Acquiring that much hardware and energy would cost an enormous amount, the attack would be visible, and it would likely crash the value of the coins and equipment the attacker owns. Waiting for more confirmations raises the bar further.

The cost of an attack

To beat the honest network, an attacker would need more hardware than all other miners combined, plus the electricity to run it, at a scale of hundreds of exahashes. Most suitable ASICs are already deployed by existing miners, so simply buying enough machines isn't possible quickly.

Defending as a receiver

• Wait for more confirmations on large payments.
• Watch for unusual reorgs; exchanges raise confirmation requirements when they see them.