Crypto Security: Threats, Defences, and Recovery

Most people picture a crypto compromise as a hacker "breaking into the blockchain." That is almost never what happens. Public blockchains such as Bitcoin and Ethereum are secured by cryptography and thousands of independent nodes; the ledger itself is not the soft target. The soft target is you, and specifically two things you control: your private keys (and the recovery phrase they derive from) and the approvals you grant to smart contracts when you use decentralised applications.

This is the threat model that matters. An attacker does not need to defeat elliptic-curve cryptography if they can persuade you to type your recovery phrase into a fake page, sign a message that hands over spending permission, or send funds to an address you believe is your own. Every attack in this reference exploits the gap between what a person thinks they are doing and what the transaction or signature actually authorises. On a public ledger, a valid signature is final: there is no bank to call, no chargeback, and no undo. Confirmed transactions cannot be reversed.

That finality is why crypto security is a discipline of prevention rather than remedy. It rewards a small number of durable habits: keeping the keys that guard significant funds offline, reading what you are about to sign, treating unsolicited contact as hostile by default, and revoking permissions you no longer use. This page catalogues the ten attack patterns that account for the overwhelming majority of self-custody losses, explains the mechanism behind each, names the single tell that most reliably exposes it, and sets out a defence stack you can actually maintain. It closes with what to do in the first hour of a compromise, and a warning about the "recovery" industry that preys on victims a second time. Nothing here is investment advice; it is operational security for people who already hold crypto.

Two ideas run through everything below and are worth fixing in mind before the specifics. First, the distinction between custody and self-custody. When you hold crypto on an exchange, the exchange holds the keys; your risk is largely their security and your account credentials, which is why SMS interception and account takeover dominate that side. When you self-custody, you hold the keys, and the risk shifts entirely to how you protect the seed phrase and how carefully you sign. Second, the distinction between an on-chain transaction and an off-chain signature. A transaction is broadcast, costs gas, and moves state immediately. A signature is a piece of signed data that may do nothing — or may be a pre-authorisation someone stores and uses against you later. Attackers exploit the fact that most people treat the second as harmless because it is free. Keep both distinctions in view and the mechanisms below become far easier to read.

The most common crypto attacks, and how each works

The attacks below are ordered roughly from the most common on-chain drains to the longer social cons. They overlap in practice — a fake support agent may steer you to a drainer site, a drainer may harvest a permit signature — so read them as a related family rather than isolated cases. For each, note the tell: the one observable detail that gives the attack away before you lose anything.

1. Wallet drainers

How it starts. You are lured to a convincing website — a fake token airdrop, a "free mint," a copycat of a real dApp, or a link posted in a compromised Discord or X account. The site asks you to "connect wallet" and then to sign a single request to claim your reward.

What actually happens. A drainer is a phishing kit, not malware. It cannot move your funds on its own; it needs you to approve one transaction or signature. When you connect, the script scans your wallet for the most valuable assets and crafts the request that extracts them most efficiently — often an unlimited token approval or an off-chain permit signature. The moment you approve, assets are swept to the attacker. Many drainers are sold as "drainer-as-a-service" kits: one developer builds the tooling and rents it to affiliates for a cut of what they steal, which is why the same fake-airdrop pattern appears across thousands of sites.

The tell. You are being asked to sign in order to receive something free. Legitimate airdrops do not require you to approve spending or sign an opaque message to "claim." A request to sign that you did not initiate from a site you trust is the signal to stop.

How to defend. Never connect a wallet holding real value to a site you reached from a DM, ad, or unsolicited link. Search results and paid ads are routinely poisoned with drainer clones of popular dApps, so type or bookmark the real URL rather than searching for it each time. Use a separate low-value "burner" wallet for mints and claims, so a bad signature can only reach what little that wallet holds. Read every signing prompt (see the defence stack below), and see how wallet drainers work for a full walk-through.

2. Token-approval abuse

How it starts. To trade or stake an ERC-20 token on a decentralised exchange, you first grant the contract permission to move that token on your behalf. This is a normal, necessary step — the approve function in the ERC-20 standard. The abuse happens when the amount approved, or the contract approved, is not what you assumed.

What actually happens. An approval sets an allowance: how much of a given token a specific contract may transfer from your address. Many interfaces default to an unlimited allowance so you never have to re-approve. That approval stays live on-chain forever until you revoke it. If the contract is malicious, or is later exploited, it can pull up to the approved amount at any time — days or months after you forgot the interaction. For NFTs, the equivalent is setApprovalForAll, which grants control over every item in a collection at once.

The tell. An approval request for an unlimited or suspiciously large amount, or a setApprovalForAll on a contract you do not recognise. The permission is broader than the single action you are trying to take.

How to defend. Prefer approving only the amount you need rather than accepting an unlimited default, even though it means re-approving next time. Audit your live approvals regularly with a tool such as Revoke.cash or the Etherscan token-approval checker, and revoke anything you no longer use — especially collection-wide NFT approvals and any allowance on a contract you cannot identify. Revoking is itself an on-chain transaction and costs gas, so batch it into a periodic clean-up rather than doing it one at a time. Full steps are in how to revoke token approvals.

3. Signature and permit phishing

How it starts. A site asks you to "sign to log in" or "sign to verify," and the prompt looks harmless because it does not cost gas. Off-chain signatures feel free and low-stakes, which is exactly what the attacker is counting on.

What actually happens. The EIP-2612 permit extension lets you approve token spending with an off-chain signed message instead of an on-chain transaction. The signature is a gasless authorisation; an attacker (or their relayer) submits it on-chain later to grant themselves an allowance, then drains the tokens — no second confirmation from you. A related trick abuses eth_sign, a low-level request that can sign arbitrary data, including a transaction, while showing you only an unreadable hash. Because no gas is spent at signing time, nothing warns you that money is at stake.

The tell. A signature request that grants spending or approval rather than merely proving who you are. Wallets increasingly decode these: watch for the words "Permit," "approve," or "setApprovalForAll," a spender address, and a token amount inside the message. A blind-signing warning is a hard stop.

How to defend. Learn to read the dangerous message types and reject anything you cannot decode into plain terms — which token, which spender, how much. A modern wallet that decodes typed-data (EIP-712) requests and warns on blind signing and eth_sign is worth using specifically for this reason. Ask why a site needs a signature at all: proving ownership for a login needs only a plain human-readable message, never a permit or an approval. See signature phishing: permit, approve and setApprovalForAll.

4. Seed-phrase theft

How it starts. A pop-up, email, fake wallet screen, or "support" agent asks you to enter, "verify," "restore," or "synchronise" your recovery phrase — the 12 or 24 words generated when you set up a self-custody wallet.

What actually happens. The recovery phrase (a BIP-39 seed phrase) is the master secret from which every private key and account in the wallet is derived. Anyone who obtains it gains total, permanent control of the wallet on any device, with no further step required. Unlike an approval, there is nothing to revoke; the only response is to move funds to a brand-new wallet before the thief does. Fake pages capture the words the instant you type them.

The tell. Anyone or anything asking for your seed phrase, for any reason. No legitimate wallet, exchange, or support team ever needs it. There is no genuine "verification," "migration," or "sync" that requires you to type it into a website or share it in a chat.

How to defend. Write the phrase on paper (or stamp it into steel) during setup, store it offline, and never digitise it — no photos, cloud notes, password managers, or email, all of which are reachable by malware or a breached account. Treat any request for it as proof of a scam, full stop. A hardware wallet keeps the seed and keys on a device that never exposes them to your computer, which is the strongest structural defence; for larger balances, an optional passphrase (a "25th word") adds a secret that is never stored on the device at all. Confirm the exact recovery-phrase policy only in the manufacturer's official documentation.

5. Address poisoning

How it starts. You notice a small or zero-value transfer in your wallet history from an address that looks almost identical to one you use often — same first and last few characters.

What actually happens. Attackers generate a "vanity" address whose leading and trailing characters match an address you transact with, then plant it in your history. On Ethereum they can do this with a zero-value ERC-20 transfer: the transferFrom function emits a valid Transfer event without checking balances or approvals when the amount is zero, so anyone can spoof a transfer from any address for the cost of gas. Later, when you copy a recent address from your history to send funds, you grab the lookalike — because people verify addresses by the first and last characters, not the full string — and your payment goes straight to the attacker. No signature of yours is compromised; you simply pay the wrong person.

The tell. A recent transaction from an address that resembles a familiar one, often with a tiny or zero amount you did not expect. The full middle of the address differs.

How to defend. Never copy send addresses from transaction history. Use a saved, verified address book so you reuse a string you once checked in full, verify the entire address rather than only the first and last few characters, and send a small test amount first for any large transfer. Where a chain or service supports human-readable names, prefer those over raw hex. Do not interact with unexpected dust; engaging with a poisoned token can invite follow-on scams. More in address poisoning explained and how to verify a contract address.

6. Fake support

How it starts. You post publicly that a transaction is stuck or a wallet is misbehaving, and a helpful "support agent" replies or direct-messages you. Alternatively, an email or pop-up branded as your wallet or exchange invites you to contact support.

What actually happens. Bots monitor social platforms for keywords such as "MetaMask help" or "wallet error" and pounce on anyone who signals confusion. The impersonator, styled to look official, already knows your problem because you posted it, so they seem credible. They then talk you into revealing your recovery phrase, entering it into a fake "validation" tool, or connecting to a drainer site to "resolve" the issue.

The tell. Support that contacts you first. Real wallet and exchange teams do not DM users out of the blue, and they never ask for your seed phrase or private keys. Unsolicited help is the red flag, not the specific story attached to it.

How to defend. Only ever start support through the official app or the verified domain you typed yourself; ignore inbound offers of help entirely, however well-timed. Never share your recovery phrase, never screen-share a session that shows your wallet, and never install "support" or remote-access software at anyone's request. When you have a problem, resist posting the specifics publicly, since that is the bait scam bots watch for. See the anatomy of fake support scams.

7. Pig butchering

How it starts. A stranger reaches you through a dating app, social media, or a seemingly misdirected "wrong number" text. The conversation is warm and patient, with no mention of money for days or weeks.

What actually happens. Named from the Mandarin sha zhu pan — "pig-butchering" — this is a long-horizon confidence fraud run at industrial scale, often by trafficked workers in scam compounds. The three phases are to find a target, "fatten" them by building trust or romance, then "slaughter" by steering them into a fake cryptocurrency investment or trading platform. The platform shows fabricated gains to encourage larger deposits; small early withdrawals may even be allowed to build confidence. When the victim tries to withdraw a large balance, they are told to pay "taxes" or "fees" first — more money in, nothing ever out.

The tell. A new online friend or partner who, sooner or later, guides you toward an investment app, exclusive opportunity, or "can't-lose" trading tip. Trust built quickly online plus a money opportunity is the signature.

How to defend. Never mix relationships formed online with investing, and never use a trading platform or "opportunity" that someone introduces you to, no matter how much rapport has built up. A demand to pay taxes or fees before you can withdraw is a defining feature of the fraud, not a bureaucratic hurdle. Reported gains you cannot freely withdraw are fake by definition. Financial-crime authorities including FinCEN have published detailed alerts on the pattern; read how the pig-butchering long con works.

8. Honeypot tokens

How it starts. You find a new token with a rising chart, active social channels, and apparent liquidity. You can buy it easily. The problem only appears when you try to sell.

What actually happens. A honeypot is a token whose smart contract is written so that most addresses can buy but cannot sell. The contract may override transfer logic to block resales, blacklist buyers silently, or impose a near-total sell fee — leaving the deployer's wallet as the only one that can cash out. Buy-only pressure inflates the price while the trap is set; the operator then removes liquidity, and every holder is left with tokens they can never convert. Your loss is the money you spent buying in.

The tell. You can buy but you cannot sell — or the contract's sell path is restricted. Hype with no verified, audited, open contract behind it points the same way.

How to defend. Treat any unaudited token pushed by social hype as hostile until proven otherwise. Inspect the verified source on a block explorer, look at the holder distribution and whether ordinary addresses have successfully sold, and be wary where liquidity or contract ownership is not locked or renounced. Automated honeypot checkers can flag obvious traps, but they are not infallible and a passing score is not a guarantee of safety. See how to spot a honeypot token and how to verify a contract address.

9. SIM swap

How it starts. Your phone unexpectedly loses signal or shows "no service," or you get a notice that your number was ported. Shortly after, accounts start receiving password resets.

What actually happens. An attacker convinces your mobile carrier — through social engineering, bribery of an insider, or stolen personal data — to move your phone number to a SIM they control. Once they own the number, they intercept SMS one-time codes and use them to reset passwords and defeat text-based two-factor authentication on email, exchange accounts, and anything else guarded by SMS. This targets custodial accounts and exchange logins rather than a properly self-custodied seed phrase, but it can be devastating where SMS is the last line of defence. The FBI's Internet Crime Complaint Center has issued public warnings about SIM-swapping used to drain financial and crypto accounts.

The tell. Sudden, unexplained loss of cellular service, or an unrequested "SIM change" or "port" notification. Your phone number is the thing being stolen.

How to defend. Move every account off SMS two-factor authentication and onto an authenticator app (TOTP) or, better, a hardware security key (FIDO2/passkey), which is not tied to your phone number and cannot be ported away. Add a carrier port-freeze or account PIN, keep the email address behind your exchange logins on the same strong 2FA, and never publish your phone number alongside your crypto identity. True self-custody sidesteps this class of attack entirely, because there is no account for a phone number to unlock.

10. Fake and malicious wallet apps and extensions

How it starts. You install a wallet from a search-ad link, a DM, an unofficial store, or even a listing that slipped past review on an official app store or browser add-on marketplace.

What actually happens. The counterfeit wallet mimics a well-known brand's interface and captures your recovery phrase the moment you create or import a wallet — either by hooking the phrase-entry screen or by serving a fake "verification" step that asks you to type your mnemonic. Some malicious browser extensions exfiltrate the seed covertly, even encoding it into innocuous-looking blockchain transactions. Researchers repeatedly find such fakes impersonating major wallets on mainstream app stores and extension marketplaces, so a listing's presence in an official store is not proof it is safe.

The tell. A wallet you installed from an ad, a message, search results you did not scrutinise, or any source other than the developer's verified site. Any wallet that asks you to enter an existing seed phrase into an unexpected screen deserves suspicion.

How to defend. Install wallets only from links on the developer's official domain, verify publisher names and download counts, and be sceptical of a "wallet update" delivered by any route other than the app's own updater. Keep the number of browser extensions with wallet access to a minimum, since any one of them can, in principle, read what happens in your browser. For meaningful balances, use a hardware wallet so the seed never touches software at all, and prefer a clean device or profile dedicated to high-value activity.

Your defence stack

No single product makes you safe. Security comes from layering a few controls so that any one mistake is survivable. The goal is not perfection but resilience: an attacker who gets past one layer still meets another, and the value they can reach at any moment is bounded. The four layers below, roughly in order of impact, cover the attacks above. A useful way to apply them is to separate your holdings by tier — a small hot wallet for day-to-day activity, and a hardware-secured wallet that rarely connects to anything for long-term storage — so that everyday convenience never puts your entire balance in the blast radius of one bad click.

1. Keep serious funds on a hardware wallet. A hardware wallet generates and stores your private keys and seed on a dedicated device that signs transactions internally and never exposes the keys to your internet-connected computer. This neutralises seed-phrase theft by malware and fake apps, and it forces you to physically confirm each action on the device screen. It does not protect you if you approve a malicious transaction — the device will happily sign what you tell it to — which is why the remaining layers still matter.

2. Practise approval hygiene. Every allowance you grant is standing risk. Approve the minimum amount where you can, review your live approvals on a schedule, and revoke anything unused — starting with unlimited allowances and collection-wide NFT approvals. Free tools such as Revoke.cash and the Etherscan token-approval checker list what each contract can currently pull from your address.

3. Verify before you sign. Slow down at the confirmation screen. Know what each request type does before approving it; the table below is the core of signing safety.

Read the spender address, token, and amount on every approval or permit, and confirm the transaction on your hardware device's own screen rather than trusting what the computer displays — malware can lie about the destination on the desktop but not on the device. If your wallet warns that it cannot decode a message, or flags blind signing, reject it. When in doubt, disconnect and start again from a URL you typed yourself; a legitimate action will still be there a minute later.

4. Keep phishing discipline. Assume unsolicited contact is hostile. Reach services only through URLs and apps you navigate to yourself, bookmark the real domains, and never act on links in DMs, ads, or emails. Use a burner wallet for mints and experiments so your main holdings are never connected to unfamiliar sites. The table below summarises where each layer helps most.

If you're compromised right now

Speed matters, but so does not making things worse. Work through these steps in order. If funds are still moving or an approval is still live, act on the highest-value assets first.

1. Disconnect and go offline. Disconnect the wallet from any dApp, close the malicious site, and if you suspect malware, take the device off the network. Do not sign anything else.
2. Move remaining assets to a new, clean wallet. If your seed phrase may be exposed, every account under it is compromised. Generate a brand-new wallet on a trusted device (ideally hardware), with a new seed, and transfer whatever remains. Do not reuse the old seed.
3. Revoke active approvals. If the loss came from an approval or permit rather than the seed, revoke the malicious allowance immediately using Revoke.cash or the Etherscan token-approval checker, and check for other live approvals.
4. Secure connected accounts. Change passwords, move two-factor authentication off SMS, and lock down the email and exchange accounts linked to your crypto. If a phone number was involved, contact your carrier about a SIM swap.
5. Document everything. Record transaction hashes, the malicious contract and recipient addresses, timestamps, the site or message that started it, and screenshots. This is what any legitimate report or investigation will need, and gathering it early — before pages disappear — is far easier than reconstructing it later.
6. Report it. File with the relevant authority in your country — for example the FBI's Internet Crime Complaint Center (IC3) in the United States, the FTC for consumer fraud, or Action Fraud in the United Kingdom — and notify the exchange if stolen funds are heading to a platform that can freeze them.

Be realistic: on-chain transactions are irreversible, and recovering stolen crypto is rare. The priority is to contain the loss and protect what remains, then to harden the setup so the same door cannot be used twice — a fresh seed on a hardware wallet, SMS removed from every account, and a review of which apps still hold approvals. A detailed, step-by-step checklist is at what to do if your wallet is compromised.

Recovery services are a second scam

In the aftermath of a loss, you will likely be approached by someone promising to get your money back. Assume they are lying. "Crypto recovery" is itself one of the most common frauds targeting people who have already been robbed, precisely because victims are motivated and easy to find — sometimes contacted using details from the original scam.

The mechanics are classic advance-fee fraud: the "recovery agent" claims to have located your funds or to have special access, then demands an up-front payment — framed as a fee, a "tax," an "AML certificate," a "gas" cost, or an "unlocking" charge — before anything can be released. Pay, and either the demands escalate or the contact vanishes. Some operations impersonate law firms, blockchain-forensics companies, or even government agencies to appear legitimate. Regulators including the FBI's IC3 and the FTC have repeatedly warned that these schemes are widespread and that agencies do not cold-contact victims to offer fund recovery.

Anyone who contacts you first, guarantees recovery, or demands a fee paid in crypto, wire, or gift cards to get your money back is running a scam. No legitimate party works that way.

Legitimate paths exist, and none of them ask you to pay a stranger up front: report to law enforcement, notify the exchange that received the funds, and if amounts are significant, consult a licensed attorney you found and vetted independently. Blockchain analytics can sometimes trace where funds moved, but tracing is not recovery — knowing that stolen crypto sits at a particular address does nothing on its own to return it, and only a lawful process against an identifiable party or a cooperating exchange can. Be especially wary of anyone who found you, references your specific case in an unsolicited message, or asks you to install remote-access software or reconnect the compromised wallet "to verify the balance"; that is a fresh attack wearing a helpful face. Full detail is in why crypto recovery services are a second scam. For definitions of any term used above, see the glossary.

Sources

• Ethereum.org — Security and scam prevention
• EIP-20: Token Standard (approve / allowance)
• EIP-2612: Permit Extension for EIP-20 Signed Approvals
• EIP-721: Non-Fungible Token Standard (setApprovalForAll)
• MetaMask Help Center — What is a token approval?
• MetaMask Help Center — Secret Recovery Phrase and private keys
• MetaMask Help Center — Spoofing and fake support scams
• Ledger Academy — Ethereum token approvals explained
• Trezor — Address poisoning attacks and how to avoid them
• Chainalysis — Anatomy of an address poisoning scam
• Revoke.cash — Check and revoke token approvals
• Etherscan — Token Approval Checker
• FBI IC3 — Criminals increasing SIM swap schemes (PSA)
• FinCEN — Alert on pig butchering (sha zhu pan) investment fraud
• FTC — What to know about cryptocurrency and scams
• FBI Internet Crime Complaint Center (IC3) — reporting