Replay Attack
A replay attack is when a valid, already-signed transaction or message is copied and broadcast again, usually on a different blockchain, so it executes a second time without the owner intending it.
A blockchain accepts a transaction if its digital signature is valid and its inputs or nonce check out. When a hard fork splits one chain into two, both chains share the same history, addresses and balances up to the split, and often the same transaction format. A transaction signed for one chain can then also be valid on the other, and anyone who sees it can rebroadcast it there.
For example, Alice holds coins on both chains after a split. She sends 5 coins to an exchange on chain A. If the chains lack replay protection, anyone can submit the same signed transaction on chain B, moving 5 of her chain-B coins to the same address. The recipient may not support that chain or may not credit or return the coins.
The main defence is replay protection. The strongest form changes how transactions are signed so a signature is valid on only one chain. Bitcoin Cash added a fork ID to its signature hash, and Ethereum includes a chain ID in signed transactions. Nonces stop the same transaction from being processed twice on a single chain. Where protection is missing, users can split their coins by spending an input that exists on only one chain.
A common misconception is that the attacker needs the victim's private key. They do not: signed transactions are public, and the attacker only copies one. Soft forks generally do not create this risk because no lasting second chain exists. Similar replays can affect smart contracts that accept signed messages without a nonce or chain identifier.