Crypto security: the complete guide

Why crypto security is a different discipline

Securing cryptocurrency is not the same problem as securing a bank account or an email inbox. In traditional finance, a fraudulent transfer can often be reversed, a stolen card can be cancelled, and a forgotten password can be reset by proving your identity to a company that holds your funds. In self-custodial crypto, none of those safety nets exist by default. A blockchain transaction that has been signed with a valid private key and confirmed by the network is treated as authoritative and is, for practical purposes, irreversible. There is no support line that can claw the funds back, because no company is holding them.

This design is deliberate. Blockchains replace a trusted intermediary with cryptography and consensus rules, which is what makes permissionless, global value transfer possible. If you want to understand the machinery underneath, our companion guide at /how-blockchains-work/ walks through keys, transactions, and consensus. The security consequence of that machinery is simple to state and hard to internalise: whoever controls the private key controls the funds, and control cannot be undone after the fact. Security therefore has to be preventive rather than reactive. Nearly every catastrophic loss in this space traces back to one of a small number of failure modes, and almost all of them are avoidable with disciplined habits.

Start with a threat model

A threat model is a plain-language description of what you are protecting, who might try to take it, and how. You cannot defend against everything at once, and trying to do so usually produces a setup that is either unusable or full of gaps. Instead, enumerate the realistic risks in order of likelihood and impact.

For most individuals the dominant threats, roughly in order, are: social engineering and phishing (you are tricked into signing something or revealing a secret); malicious token approvals and drainer contracts (you grant a contract permission to move your assets); malware and clipboard hijackers on your own device; physical compromise or loss of a device or a written seed phrase; and custodial failure if you keep funds on an exchange that is hacked, becomes insolvent, or freezes withdrawals. Nation-state attackers and targeted supply-chain compromises exist but are not the primary concern for typical users, whereas an unsolicited direct message offering “support” almost certainly is.

The value of writing this down is that it lets you match effort to risk. Someone holding a small balance for occasional transactions has a very different profile from someone self-custodying a life-changing sum. The rest of this guide is organised around the high-likelihood threats, because that is where disciplined habits produce the largest reduction in real risk.

Private keys and seed phrases: the root of everything

Underneath every wallet is a private key: a very large secret number. From it, cryptography derives your public key and your public address, and only the private key can produce the digital signatures that authorise spending. Modern wallets rarely ask you to handle a raw private key directly. Instead they show you a seed phrase (also called a recovery phrase or mnemonic): an ordered list of common words, usually twelve or twenty-four of them.

That phrase is not a password and it is not a hint. Under the widely adopted BIP-39 standard, the words are drawn from a fixed list of 2,048 words, and the ordered sequence encodes the entropy from which every key in your wallet is deterministically derived. Because the same phrase always regenerates the same keys, your seed phrase is your wallet. Anyone who reads it can reconstruct your keys on any compatible wallet, anywhere, with no PIN and no device. The mechanics of how a phrase becomes a hierarchy of keys are covered in /wallets-and-custody/; the security rule that follows from them is absolute:

• No legitimate wallet, exchange, “support agent”, airdrop, or verification process will ever need your seed phrase. A request for it is, without exception, an attempt to steal your funds.
• Never type your seed phrase into a website, a chat, an email, a note-taking app, a password manager field labelled “secure note”, a photo, or a cloud document. The moment a phrase touches an internet-connected system, you must assume it can be exposed.
• Treat the phrase as the master credential it is, distinct from the day-to-day PIN or password that merely unlocks a specific device.

Backups and storage that survive both attackers and accidents

Seed-phrase handling has two opposing failure modes, and a good backup defends against both. If the phrase leaks, you are robbed. If the phrase is lost or destroyed and you have no other copy, the funds are gone just as permanently, because no one can regenerate the keys for you. The goal is redundancy without exposure.

Sound practice is to record the phrase on a durable physical medium, kept offline. Writing it clearly on paper is the baseline; metal backup plates that resist fire and water are a common upgrade for larger holdings. Store at least one backup in a second physical location so that a single fire, flood, or theft does not wipe out both the funds and the ability to recover them. Keep the copies away from obvious targets like a desk drawer next to the computer. Avoid ever creating a digital copy: no screenshots, no cloud notes, no emailing it to yourself. If you want an additional secret that even a person who finds the written words cannot use, an optional BIP-39 passphrase (sometimes called the “25th word”) adds a component that is never stored with the phrase; its trade-offs are detailed in /wallets-and-custody/. Whatever scheme you choose, test that you can actually restore from your backup before you rely on it, ideally with a trivial amount first.

Token approvals and allowances: the permission you forget you gave

A large share of losses on smart-contract platforms such as Ethereum do not involve a stolen seed phrase at all. They exploit token approvals. On these networks, most tokens follow the ERC-20 standard, and to let an application move your tokens you must first call an approve function that sets an allowance: the amount a given contract address is permitted to spend on your behalf. The application then uses transferFrom to move up to that allowance.

For convenience, many interfaces request an unlimited allowance (technically a value near the maximum a 256-bit number can hold) so you never have to approve again. The security problem is that the approval persists indefinitely and independently of your original transaction. If the contract you approved is malicious, later found to be vulnerable, or is upgraded by a compromised team, the outstanding allowance lets it drain the approved token from your wallet at any time, without asking again. For NFTs and similar assets, the analogous danger is setApprovalForAll, which can grant a contract control over an entire collection in a single signature.

The defensive habits are concrete. Prefer approving only the amount you intend to use rather than an unlimited allowance when the interface offers the choice. Periodically review and revoke allowances you no longer need, setting them back to zero. Independent tools such as Revoke.cash and the token-approval views built into block explorers let you audit and reset outstanding approvals. Because this topic is where careful users still get caught, we cover the mechanics and the step-by-step process in /how-to-revoke-token-approvals/. A newer pattern, Uniswap’s Permit2, lets applications rely on signed, off-chain, time-limited permits instead of separate on-chain approvals; it is convenient but means a single signature can carry spending authority, so the same principle applies: understand what a signature grants before you produce it.

How wallet drainers actually work

“Drainer” refers to a category of malicious toolkit engineered to empty a wallet as soon as the victim connects and signs. Drainers are sold as ready-made kits and deployed behind convincing front-ends, which is why they succeed against otherwise careful people. The attack rarely needs your seed phrase. Instead it manufactures a moment where you willingly sign something whose true effect is hidden.

Common vectors include fake versions of well-known applications reached through sponsored search ads or lookalike domains; fraudulent “airdrop claim”, “token migration”, or “wallet validation” pages; compromised project accounts posting malicious links; and direct messages impersonating support. Once you connect, the drainer presents a transaction or an off-chain signature that grants a broad token approval, executes a transferFrom, or authorises a Permit2 transfer. Because the request may be displayed as opaque hexadecimal data rather than a human-readable summary, users approve it without realising it hands over control. Signing blind is the single most dangerous habit in this space. A full anatomy, with annotated examples of what the malicious prompts look like, is at /how-wallet-drainers-work/.

Two defences neutralise most drainers. First, read what you sign. Modern wallets increasingly decode transactions into plain language (“Approve unlimited USDC to 0x…”); if the effect is unclear or the wallet warns you, stop. Second, compartmentalise. Use a fresh or low-value “hot” wallet for interacting with unfamiliar applications and mints, and keep long-term holdings in a separate wallet that never touches those sites. If a drainer catches your interaction wallet, the loss is bounded.

Phishing, impersonation, and address poisoning

Phishing in crypto extends far beyond fake login pages. Because there is no password to reset, attackers aim either to extract your seed phrase or to obtain a malicious signature. Recognising the recurring patterns is more useful than memorising specific scams.

Impersonation and urgency. Unsolicited “support” contact, especially by direct message, is a near-universal red flag; legitimate teams do not DM you first and do not ask for secrets. Attackers manufacture urgency (“your wallet is compromised, migrate now”) to short-circuit caution. Slow down; urgency is a tool used against you.

Fake domains and search ads. Attackers buy ads and register lookalike domains for popular applications. Reach important sites through your own saved bookmarks rather than search results or links in messages, and verify the domain character by character.

Address poisoning. This attack exploits the habit of copying an address from your transaction history. An attacker watches your on-chain activity, generates a “vanity” address whose first and last characters match one you recently transacted with, and sends you a tiny or zero-value transfer so their lookalike address appears in your history. Later, you copy the wrong one and send funds to the attacker. Reported losses from this technique have been substantial, including a widely reported case in which a user lost tens of millions of dollars in wrapped Bitcoin after copying a poisoned address. Defend by never copying addresses from transaction history, verifying the full address rather than only the ends, and using an address book or verified name service entries. Wallet vendors including MetaMask, Ledger, and Trezor document this pattern in their help centres.

Transaction hygiene: verify before you sign

Because signing is the irreversible act, disciplined verification at signing time prevents most losses even when everything upstream has gone wrong. Adopt a fixed routine and apply it every time, regardless of how routine the action feels:

1. Confirm the destination. Verify the full recipient address, not just the leading and trailing characters. Where possible, send a negligible test amount first for large or first-time transfers.
2. Understand the action. Read the decoded transaction. Is it a transfer, an approval, or a contract call? An unexpected approval request during what should be a simple swap is a warning sign.
3. Check the amount and the allowance. Be suspicious of unlimited approvals and of amounts that do not match what you intended.
4. Verify on the device. If you use a hardware wallet, confirm that the address and amount shown on the device’s own screen match what your computer or phone displays. Malware can alter what the computer shows but cannot alter the hardware screen. This on-device verification, explained further in /wallets-and-custody/, is one of the strongest protections available.
5. Never sign what you cannot read. If a prompt is opaque and you cannot establish its effect, decline.

Incident response: what to do if you are compromised

Speed and correct sequencing matter enormously if you suspect compromise, but panic causes mistakes. The right first action depends on how the compromise occurred, and it is important not to make things worse.

If your seed phrase may be exposed (you entered it somewhere, it was photographed, or a device holding it was compromised), assume every account derived from that phrase is fully compromised. Revoking approvals will not help, because the attacker can sign anything. Immediately move remaining assets to a new wallet generated from a fresh seed phrase on a clean, trusted device. If an attacker is actively draining, prioritise the highest-value and most-liquid assets first; some assets may not be saveable, and that is an unavoidable consequence of a leaked master key.

If you signed a malicious approval but your seed phrase is intact, the attacker can only move what you approved. Revoke the specific allowance immediately using a tool like Revoke.cash, and move at-risk tokens to safety. Then audit all outstanding approvals for other malicious grants.

In all cases: disconnect the affected device from the internet if malware is suspected, document transaction hashes and addresses for any report, and be acutely aware that “recovery services” promising to retrieve stolen crypto are themselves overwhelmingly a follow-on scam. A structured, step-by-step checklist is maintained at /wallet-compromised-what-to-do/. You can report incidents to authorities such as the FBI’s Internet Crime Complaint Center (IC3), whose 2024 report attributed roughly 9.3 billion US dollars in losses to cryptocurrency-related fraud across nearly 150,000 complaints — a scale that underscores why prevention is the only reliable strategy.

Device and operational security

Every hot wallet inherits the security of the device it runs on, so the device itself is part of your threat surface. A phone or computer compromised by malware can steal keys, swap a copied address for the attacker’s, or manipulate what a transaction appears to do. Operational security is therefore not separate from crypto security; it is a precondition for it.

Practical measures are unglamorous but effective. Keep the operating system and browser patched, since many compromises exploit known, already-fixed vulnerabilities. Install as few browser extensions as possible and audit the ones you have, because an extension with broad permissions can read and alter pages including wallet interfaces. Be disciplined about where you download software, verifying you are on the genuine vendor domain rather than a sponsored ad or clone. For meaningful holdings, consider a dedicated device — a separate phone or a clean profile — used only for crypto and nothing else, which shrinks the surface that malware and malicious sites can reach. Beware “clipboard hijacker” malware specifically: it silently replaces a copied address with the attacker’s, which is one more reason to verify the full destination on your signing device rather than trusting a paste.

Authentication for custodial accounts deserves the same care. Prefer app-based or hardware security keys over SMS one-time codes, because phone numbers can be captured through SIM-swap attacks in which an attacker convinces a carrier to move your number to their device. Where a provider offers it, a hardware security key is the strongest widely available second factor. These measures do not touch the blockchain at all, but they close the doors through which most real-world compromises actually arrive.

A red-flags checklist

Attacks vary in detail but rhyme in structure. The following signals should each trigger a full stop and independent verification before you act, no matter how legitimate the surrounding context appears:

• Anyone asking for your seed phrase or private key, in any wording or for any stated reason.
• Unsolicited direct messages offering “support”, “verification”, “migration”, or help recovering funds.
• Urgency and pressure: “act now or lose access”, “limited time”, “your wallet is at risk”.
• A signing request you cannot read, or one that appears during an action that should not require it (for example, an approval prompt during what should be a simple connect).
• A request for an unlimited approval or setApprovalForAll where a bounded amount would suffice.
• A link reached via search ad, DM, or email rather than your own saved bookmark, or a domain that is subtly misspelled.
• An address copied from transaction history rather than a verified source.
• Any offer that pairs an unexpected windfall (free airdrop, giveaway, “you have unclaimed tokens”) with a wallet connection or signature.

None of these is proof of an attack on its own, but each is common enough in real thefts that treating it as a stop sign costs you nothing and prevents most losses.

Ongoing hygiene that compounds

Security is a set of habits, not a one-time configuration. A few practices, applied consistently, disproportionately reduce risk over time.

• Tiered custody. Keep long-term savings in cold storage (a hardware wallet, or multisig for larger amounts) and only a spending balance in a hot wallet. See /wallets-and-custody/ for how to structure this.
• Separate wallets by purpose. One wallet for holding, another for active dApp interaction, so that a bad signature cannot reach your reserves.
• Keep software genuine and current. Download wallets and firmware only from official sources, and apply updates that patch known vulnerabilities. Verify you are on the correct application rather than a clone.
• Reduce your on-chain footprint where practical, and be mindful that public addresses expose your balances and history to anyone, which is what enables targeted attacks like address poisoning.
• Periodically re-audit approvals and revoke anything you are no longer using.
• Use exchanges deliberately. Funds on a custodial exchange are exposed to that company’s solvency and security; withdraw to self-custody what you are not actively trading, and enable the strongest available account protections, preferring app-based or hardware authentication over SMS.

None of these habits is difficult in isolation. Their power comes from consistency: an attacker needs you to slip only once, so the objective is to build routines that hold up even when you are tired, rushed, or excited about a new opportunity. For definitions of the terms used throughout this guide, see the /glossary/.

Sources

• ethereum.org — Staying safe: common scams and security
• MetaMask Help Center — Address poisoning scams
• Trezor — What are address poisoning attacks and how to avoid them
• Ledger Academy — Ethereum token approvals explained
• Revoke.cash — Learn: token approvals and allowances
• FBI IC3 — 2024 Internet Crime Report (PDF)
• FBI — Annual Internet Crime Report press release
• BIP-39 — Mnemonic code specification