
Two blockchains cannot read each other. When you move ETH or a token from Ethereum to another network, the destination chain has no native way to see that you locked funds at the source. Something has to tell it. Every bridge design is a different answer to who that something is, and every major bridge loss traces back to that answer.
Why bridges concentrate risk
Most bridges lock the original asset in a contract on one chain and issue a representation on the other. That has three consequences. The locked collateral sits in one place, so the value at risk is the whole pool rather than any single user’s balance. Some privileged component decides when to release it. And the two chains share no source of truth, so correctness depends on the bridge’s own attestation. Ethereum’s developer documentation on bridges lists the resulting risks plainly: smart contract risk, systemic risk from wrapped assets, and counterparty risk where users must assume “validators will not collude to steal user funds”.
Four ways to attest
External validator or multisig sets
A group of signers watches the source chain and signs when a deposit occurs. The trust assumption is explicit: a threshold of signers is honest and their keys are safe. The important number is not the nominal threshold but how many genuinely independent parties hold the keys. Fifteen signers run by three organisations on one cloud provider is a three-party assumption.
Light-client and proof-based bridges
The destination chain verifies the source chain’s consensus or state directly, so the assumption reduces to the security of the source chain. This is the strongest design and the most expensive, because verifying another chain’s consensus on-chain costs gas and must be built separately for each pair. Zero-knowledge proofs of consensus are reducing that cost. These bridges still have upgrade keys.
Optimistic bridges
Messages are assumed valid unless challenged within a window. The assumption is that at least one honest watcher exists and can act in time. The cost is delay, and the practical question is whether anyone is still watching during a quiet year.
Liquidity networks
Nothing is minted. A provider on the destination chain pays you from its own inventory and settles with its counterpart later. This removes the catastrophic failure of unbacked minting; the worst case is usually a stuck or failed transfer rather than a wrapped token that loses its backing.
Canonical rollup bridges are a special case
The bridge built into a rollup is part of its settlement contract on Ethereum. Withdrawals are paid only after the rollup’s own proof or challenge rules are met, which is why exits from optimistic rollups take days. That makes the canonical bridge slower than third-party routes, and its trust assumptions are those of the rollup itself, including whoever can upgrade it. See our glossary entry on the canonical bridge.
Where the weak points sit
Follow the attestation path and the soft spots are the parts around the cryptography rather than the cryptography itself: how signing keys are stored and whether one compromised machine can reach the threshold, whether the verification code accepts only correctly formed messages, whether a privileged role can be reassigned, and whether an upgrade can change any of the above. Scrutiny is best spent on key management, verification code and upgrade control.
The Settlement Test
- What settles: the lock or burn on the source chain, and the mint or release on the destination chain. They are two separate settlements joined only by the bridge’s attestation.
- What it costs: gas on both chains, a bridge or liquidity fee, and for canonical rollup exits the time of the challenge or proof window. Our bridge cost calculator adds up the parts you can see.
- What breaks: if the attestation is forged or the attesters collude, the destination asset can become unbacked while still trading normally, because nothing on that chain knows.
What a user can check
Who can authorise a release, and how independent are they really? Is the contract upgradeable, by whom, and with what delay? Does anyone reconcile locked collateral against circulating wrapped supply? How long has the bridge held its current value? None of these certify safety; they filter out the obviously fragile. The durable mitigation is limiting how much sits in any bridge at once. Our guide to bridges and their failure modes covers the mechanics in more depth.
Primary sources
Educational content, human-edited and possibly AI-assisted. It describes how systems work and is not financial, legal or tax advice. Report an error via our corrections page.


