Passkeys, explained: why there is nothing to phish
A passkey is a key pair, not a secret: the private key never leaves your device, so a fake login page has nothing to steal. How it works, where it breaks, and where to start.

On this page
In brief
- A passkey is a key pair, not a shared secret: the private key never leaves your device and the server holds only the public key.
- The credential is bound to the site's exact domain, so on a fake login page there is no matching credential to sign with.
- CISA calls FIDO/WebAuthn the only widely available phishing-resistant authentication; SMS codes are a last resort.
- Synced passkeys move between your devices encrypted; device-bound keys on a hardware key never leave it.
- Recovery is the weak point: keep a second device or a hardware key, and start with your email, then your password manager.
The fake login page looks exactly like your bank. Same logo, same layout, same little padlock in the address bar — or close enough that you would not notice the difference at 8 a.m. You type your username and password, then the six-digit code from your authenticator app, and the page thanks you and sends you on your way. Somewhere else, someone logs in as you with the code you just handed them.
That is the first attack in CISA's October 2022 fact sheet on phishing-resistant authentication: a page mimicking a company's legitimate login portal, collecting the username, the password, and the six-digit code in one visit. The code was supposed to be the part that saved you. It turns out codes are phishable too.
A passkey fixes this by removing the thing the fake page is built to collect. There is no password to type, no code to copy, no secret to relay. The credential never appears on the page at all.
A key pair instead of a secret
A passkey is a pair of cryptographic keys, one public and one private, created for a single website or app. The private key stays on your device (phone, computer, or hardware security key) and never leaves it. The website keeps only the public key, which is useless on its own. To sign in, the site sends your device a one-time challenge, the device signs it with the private key, and the site checks the signature against the public key it holds.
You approve the signing the same way you unlock your device: a fingerprint, a face scan, a PIN. The biometric check happens on the device and stays there. Biometric information is never sent to any remote server; the server only learns that the check passed.
That is the whole mechanism. A password is a shared secret: the site knows it, you know it, and anyone who tricks you into typing it knows it too. A passkey has no shared secret. There is nothing to leak in a breach, nothing to reuse across sites, and nothing for a fake login page to harvest.
The key only fits one door
What makes a passkey phishing-resistant, rather than merely convenient, is that the credential is bound to the site it was created for. The WebAuthn standard scopes each credential to the site's exact domain, and the device will only sign for that site.
So when you land on a lookalike domain, there is simply no matching credential. The browser has nothing to offer, the private key has nothing to sign, and the fake page gets nothing to relay. The protection does not depend on you spotting the fake domain. It depends on cryptography, which does not get tired.
This is why CISA's fact sheet calls FIDO/WebAuthn "the only widely available phishing-resistant authentication", and ranks every other common method below it. App-based codes and push approvals with number matching are vulnerable to phishing, even if they resist push bombing. Push approvals without number matching can be beaten by flooding you with prompts until you tap accept by accident. CISA calls this push bombing, or push fatigue. SMS and voice codes sit at the bottom, vulnerable to phishing, weaknesses in the phone network's signalling protocol (SS7), and SIM swaps, and CISA says they should be used only as a last resort. The other qualifying method, PKI smart cards like the ones government agencies use, is strong but needs mature identity management and is not widely supported by consumer services.
Synced or bound to one device
Passkeys come in two kinds, and the difference matters when you lose a phone. A synced passkey is copied between your devices through your passkey provider (iCloud Keychain, Google Password Manager, 1Password, Dashlane) over an end-to-end encrypted sync, so one created on your phone is waiting on your laptop. A device-bound passkey never leaves a single device, which is what you get on a hardware security key.
Sync is the convenient default. A hardware key is the stronger choice for the accounts that matter most, because the private key cannot be copied off it at all.
Recovery is the question every passkey explainer has to answer. If your passkeys live only on one phone and the phone is gone, you are back to the site's account recovery, which is often the weakest part of the whole system. So do not keep a single copy. Turn on sync with a provider you trust, register a passkey on a second device, or keep a hardware key as a spare. The WebAuthn specification suggests registering more than one authenticator per account, for exactly this reason.
It is already everywhere, quietly
Google said in May 2024 that passkeys had been used to authenticate people more than a billion times across over 400 million Google Accounts, and that on a daily basis they were already used more often than SMS codes and authenticator-app codes combined. Google also says they are about 50% faster than passwords. Amazon, PayPal, eBay, Uber, WhatsApp, Shopify, Docusign, Kayak, Mercari and Yahoo! JAPAN all rolled out passkey support in the year before that announcement. A 2024 survey commissioned by the FIDO Alliance found 53% of people had enabled passkeys on at least one account.
Amazon's six-times-faster claim and Yubico's 99.99% phishing-reduction claim are their own measurements, but the direction is consistent across every source that publishes one. Verizon's finding, cited by the FIDO Alliance, is that 77% of hacking-related breaches involve stolen credentials, which is the problem passkeys are sized to solve. Stolen credentials are also how most breach chains start: one recent chain ran from a breached vendor through a stolen API key to tampered extensions.
What passkeys do not do
A passkey stops credential phishing. It does not stop everything. If malware on your device steals the session cookie after you sign in, the attacker rides your session without ever touching the login. If a site does not offer passkeys, you are back to passwords and codes. And a weak account-recovery flow, an email reset link or an SMS fallback, can still be the way in. Passkeys fix the front door. They do not fix the spare key under the mat.
CISA tells organisations to start with the systems attackers target first, email, file servers, remote access, identity servers, and the highest-value accounts first. The same order works for one person. Start with your email, because a compromised inbox resets everything else. Then your password manager, then your bank. Most sites offer them under Settings, then Security, labelled "passkeys" or "sign in with a passkey". Keep the password until the passkey works, then delete it where the site lets you.
The password had a long run. It asked ordinary people to memorise dozens of unique secrets, never reuse them, and never type them into a page that looked almost right. That was never a fair ask. A passkey asks for a fingerprint and gives the fake page nothing to take. For the first time, the easy way to sign in is also the hard way to steal.


