How crypto wallets actually fail

“Which crypto wallets have been hacked” is a reasonable question built on a mistaken assumption: that the wallet is the thing that breaks. In most losses it is not. The software did exactly what it was designed to do, with a key it was entitled to use, on the instruction of someone holding the right credential.

Here are the ways funds actually leave a self-custodial wallet, ordered roughly by how much money each has cost.

The phrase was somewhere it could be read

The largest category, and the least dramatic.

A photograph in a camera roll that syncs to a cloud account. A note in a password manager whose master password was reused. A screenshot. An email to yourself. A text file on a laptop that later had something else installed on it. Each converts a secret only you hold into one held by a service that can be breached, compelled, or simply accessed by someone who learns one other password.

No wallet defends against this, because from the software’s point of view nothing went wrong: somebody knew the phrase and restored it, which is precisely what a phrase is for. What the phrase controls and how it differs from a password are the two things worth understanding before deciding where to keep it.

You authorised it

The second largest, and the one that beats experienced holders.

An approval granting a contract permission to move your tokens. A signed message that was an order rather than a login. A “claim” page whose transaction did something other than claim. The key never left, no software was compromised, and the transaction is valid — which is why it is also irreversible. What a wallet connection actually approves is the mechanism in full.

The software you installed was not the software

Supply chain, and it is the case where the wallet genuinely is implicated — though usually not the version its authors shipped.

A fake application in an app store with a plausible name and good reviews. A download from an advert above the real search result. A browser extension that changed hands and pushed a malicious update to people who had installed something legitimate months earlier. A compromised dependency in the build, which affects everyone who updates and nobody who reads the interface.

The defences are unglamorous: install from an address you navigated to yourself, prefer software whose releases can be checked against published source, and treat an unexpected permission request after an update as a reason to stop.

Weak key generation

Rare, catastrophic, and the one case where the wallet is unambiguously at fault.

Keys are only unguessable because they are drawn from a very large space at random. A wallet that generates them from a poor source of randomness — a weak platform generator, a timestamp, or a passphrase a human invented — produces keys that live in a much smaller space, and anyone can search that space at leisure. Funds sitting in such an address are not stolen so much as collected, sometimes years later, by someone sweeping every key a flawed generator could have made.

This is the strongest argument for well-reviewed, widely-used software over something novel, and for whether anyone outside can verify the build being a real criterion rather than a box.

The address you pasted was not the address you copied

Clipboard interference: malware that watches for something shaped like an address and substitutes its own. You paste, it looks approximately right, you send.

The habit that defeats it costs five seconds — check the last several characters on the confirmation screen, not in the field you pasted into, because the confirmation is rendered from what will actually be signed. A hardware device’s screen is the strongest version of the same check.

Someone made you do it

Worth stating plainly because it is the failure mode that scales with how public you are about holdings. No cryptographic property protects a key from its owner being persuaded, deceived at length by someone patient, or coerced.

The mitigations are behavioural — not discussing holdings, not being identifiable as a holder — plus, on some hardware wallets, a passphrase that opens a different account from the one a coerced disclosure would reveal.

So which wallets have been hacked

A few, genuinely: supply-chain compromises and generation flaws are real and have happened. But if you sort the losses by value, the software’s own failures are a small share, and the list of “hacked wallets” tells you much less than it appears to — it is a record of which brands had an incident, not of which arrangement is sound.

The better question is which of the failure modes above your setup is exposed to, because that is answerable about you and every item on it is something you can change.

Storing bitcoin for a decade

The long-term case deserves its own answer, since it is the one where every decision compounds.

Cold storage, so the key is never on a networked machine — the line between hot and cold is the relevant threshold. A standard BIP-39 phrase on a standard derivation path, so recovery does not depend on one vendor still existing in 2036. A restore tested on different software before anything of value is committed, because an untested backup is a belief rather than a backup. Backups on something that survives water and fire, in more than one location, since the failure over ten years is as likely to be a house fire as an attacker. And a written note for whoever needs to find it if you cannot — the most common total loss over a long horizon is nobody knowing the phrase existed.

Nothing in that list is about which device you bought.