Security · Guide 7

Smart contract and bridge risk

A contract can be audited, immutable and still wrong. The risk is not only the code: it is who can change it and what it trusts.

A smart contract can be audited, immutable and still wrong. The risk is not only the code you can read. It is who can change it, what it trusts for prices, and how much of your money sits behind a single key you will never meet.

Approvals are the permission you forget

When you interact with a token, you usually grant a contract the right to move it. That right persists after the transaction, after the tab closes and after you stop using the site. It is not a session; it is a standing instruction on a public ledger.

This is why a drained wallet can be drained twice. The signature that emptied it is still valid, so the address stays unsafe until the approval is revoked.

Unlimited allowances

Many interfaces ask for an unlimited allowance so you never have to approve again. It is convenient for the site and bad for you: the exposure is your entire balance of that token, for as long as the approval exists.

Exact amounts are better. If a site only offers unlimited, that tells you something about its priorities. Approving the exact amount is one extra transaction per use and one less permanent liability.

Revoking

Revoking is the reverse operation: you send a transaction that sets the allowance back to zero.

Review approvals periodically, and immediately after using anything new. Several wallets and independent tools show the list; the tool matters less than the habit of looking. If you cannot recognise a contract, treat it as hostile until you can.

Two details people miss. Revoking does not undo a past transfer, and revoking on one chain does nothing on another. If you used the same address on several networks, check each one.

Upgradeable contracts

Many contracts can be changed after deployment, usually through a proxy pattern: the address stays the same, the logic behind it is swapped. Upgradeability is not a flaw; it is how bugs get fixed. It is also how a safe contract becomes a hostile one overnight, with no new address to warn you.

Before trusting a contract with meaningful value, find out whether it is upgradeable and who holds the upgrade right. If the answer is «a multisig we will decentralise later», the honest translation is «a small group can change the rules today».

Admin keys

Almost every protocol has privileged functions: pause transfers, change fees, mint, blacklist, withdraw from a treasury, or point the contract at a new price source. The question is not whether these powers exist, but who holds them and what limits them.

Signals worth checking: is the key a single address or a multisig, is there a timelock between a decision and its effect, and is the key documented anywhere you can read. A timelock turns a surprise into a warning. A single key with no delay turns a compromise into a total loss.

Oracles

A contract that needs a price asks an oracle. If the oracle can be manipulated, the contract can be drained without a single bug in its own code. Thin markets are the classic setting: push the price on a small pool, borrow or liquidate against the distorted number, repeat.

For a user, the question is whether the protocol depends on a deep market for its price and whether it uses several independent sources. You will rarely be able to verify that yourself, which is a reason to size positions accordingly.

Bridges

A bridge holds assets on one chain and issues a representation on another. That design concentrates risk: the locked assets are a single prize, the validator set or multisig is a single target, and the wrapped asset depends on the bridge staying honest and solvent.

Bridges have produced some of the largest losses in the sector, and the pattern repeats across designs. Using one is a deliberate decision to trust a small group of operators and a large pile of code at the same time. Do not leave funds sitting on a bridge, and do not treat a wrapped asset as identical to the original.

Audits are not warranties

An audit is a review by a competent team at a point in time. It can find real problems and it cannot prove absence of them. It says nothing about the admin keys, the oracle, the operational security of the team, or the code added after the review.

Read the audit's scope and date, not the logo on the homepage. «Audited» next to a contract that has been upgraded since is a statement about a different contract.

The useful question is never «is this safe?». It is «what can go wrong, who can cause it, and how much am I willing to lose if they do».

What to do this week

  • Open your approval list on each chain you use and revoke anything you cannot explain.
  • For one protocol you hold funds in, find out whether it is upgradeable and who holds the admin key.
  • Decide a maximum you are willing to leave in any single contract, and move the rest out.

Next: how tokens lure you into signing in the first place, in fake airdrops and rug pulls, and how to read the request itself in phishing and wallet drainers. The mechanics behind the request are in smart contract risk: approvals, oracles and what an audit does not cover.