cryptobank
Crypto Bank is an independent editorial publication. It is not a bank, is not a deposit-taking or custodial institution, holds no banking, trust, money-transmitter or e-money licence, and offers no FDIC or FSCS protection. Nothing here is financial, legal or investment advice.
BTC $63,224-0.77% ETH $1,773-0.88% USDT $1.00+0.10% USDC $1.00+0.01% SOL $73.99+2.41% XRP $0.98+2.53% BNB $584-0.40% ADA $0.15+0.59%
The Risks No Custodian Covers: Approvals, Impersonation and Address Tricks

GUIDE

The Risks No Custodian Covers: Approvals, Impersonation and Address Tricks

Crypto Bank Editorial Team · November 18, 2026

When you move assets into self-custody, the custodian’s insurance, fraud team and recovery process leave with it. What remains are attacks designed for an audience whose errors are final. They rarely involve breaking cryptography. They involve persuading the key holder to do something: sign, paste, reveal, hurry. Recognising the recurring shapes is more reliable than inspecting every link, and a few structural habits remove whole categories regardless of how convincing a particular attempt looks.

Standing approvals: permission that outlives the moment

On networks with token allowances, you do not hand a contract your tokens; you grant it a standing permission to move up to an amount of a token from your address, whenever it chooses. Three properties surprise people. The permission persists after the transaction it was granted for. Interfaces often request the maximum possible amount, so it covers your whole balance of that token, including tokens received later. And it is granted to a contract address, not to an app; if a website is tampered with and asks you to approve a different address, nothing on the page looks wrong.

That is why the most effective attacks do not ask you to send anything. The prompt looks like connecting, claiming or signing in. No value leaves at that moment; the transfer happens later. Approvals are public on-chain data: a block explorer’s approvals view for your own address shows which spenders you have authorised, read-only, with no wallet connection needed. Anything that insists on a wallet connection before it will show you your approvals deserves suspicion.

Reviewing approvals does not protect against a leaked recovery phrase, a transfer you sign yourself, or permissions granted by off-chain signed messages, which may not appear in a standard approvals view. It is maintenance, not a security strategy.

Impersonation: support that contacts you first

Mention a problem in a public forum and a helpful “support agent” often appears in your private messages within minutes. Whatever follows (a validation site, a sync tool, a screen share) exists to reach your recovery phrase or obtain a signature. The rule that defeats this is absolute: anyone who contacts you first is not support. Data leaks make this harder to spot, because the caller may know real details about you. Coinbase’s May 2025 disclosure described stolen customer data being used in social-engineering attempts; Coinbase said private keys were not compromised, but customers were targeted with convincing pretexts (profile).

Urgency about your own security

“Your wallet is compromised, migrate now.” The message uses fear about security to push you into the action that destroys it. No legitimate wallet provider requires a migration within minutes, and none will ever need your recovery phrase. Time pressure on a security decision is itself the signal.

Addresses that look right

Two techniques exploit the habit of checking only the first and last few characters. Clipboard malware swaps a copied address for the attacker’s at the moment you paste. Address poisoning sends you a tiny transaction from an address whose ends match one you use, hoping you later copy it from your history. Counter-measures: verify the full address on the signing device’s screen, keep a trusted address book, and send a small test amount before anything substantial.

Compromised official channels

The hardest variant: a genuine project account, community server or website is taken over and posts the malicious link itself. Checks based on the sender fail. The 2023 Ledger Connect Kit incident is a supply-chain version of the same idea: by Ledger’s account, a tampered library served through legitimate applications led users to sign draining transactions. What still works is confirming announcements across several channels, treating time-limited offers with extra suspicion, and reaching sites through your own bookmarks.

Tokens you did not buy

Unexpected tokens appearing in a wallet are bait. Trying to sell or claim them leads to a site that requests an approval, and the approval is the attack. Holding them costs nothing; interacting is the risk. Hide them and move on.

Structure beats vigilance

Care runs out on a bad day; structure does not. The highest-leverage habit is separation: a holdings wallet that never connects to any website, and a separate interaction wallet with a small balance for everything else. An approval signed from the interaction wallet exposes only what is in it. Add bookmarks instead of search results (paid ads are sometimes bought for clone sites), read the device screen rather than the web page, test-send before large transfers, and review approvals when nothing is wrong, not in the twenty minutes after news of an exploit, when fake “revoke” and “recovery” sites are most active. These habits succeed against attacks that also catch experienced, careful people.

The Custody Question, applied

  • Who holds the keys: you, plus anyone holding a live approval over your tokens. Count both.
  • Who insures it: nobody. Custodian crime policies do not follow assets into self-custody, and signed transactions are not reversible.
  • What happens on default: there is no counterparty; losses from these patterns are final. Compare platform risks and what a hardware wallet protects.

Sources

Editorial explainer, not security advice for your specific setup. Crypto Bank is not a bank and does not hold client assets.