Picture Farida, an accounts executive at a logistics firm in Indore. It is month-end and she is tired. An email arrives that looks like it is from her company’s single sign-on provider: “Your session expired, sign in again to see the pending invoices.” The page it links to is a faithful copy, logo, colours, even the same error message for a wrong password. She types her password. A moment later it asks for the six-digit code from her authenticator app. She types that too.

Everything she did was correct by the usual advice. She used a strong password and a second factor. And the attacker, whose page was quietly passing her entries to the real site in real time, is now logged in as her.

Hold on to Farida. We will replay her morning with a different kind of login and see what changes.

What was wrong with the old picture

A password and a one-time code are both shared secrets typed by a human. Whatever a human can type into a page, a convincing fake page can capture and relay. The security of the whole arrangement depends on Farida noticing the address bar, which is a lot to ask at 6 p.m. on the 30th.

The idea behind FIDO2 is to take the human judgement out of the loop. It does that with public-key cryptography, which you may know from certificates: a pair of keys where the private key never leaves its owner, and the public key can be given to anyone.

How a passkey login works

FIDO2 is the umbrella name for two specifications: WebAuthn, the browser-side interface standardised by the W3C, and CTAP, the protocol by which a browser talks to an authenticator such as a security key or a phone. A passkey is the user-friendly name for a FIDO credential that can be synced or used across devices.

Registration. When Farida adds a passkey to her account, her device creates a new key pair just for that website. The public key goes to the site and is stored against her account. The private key stays in the device’s secure hardware or the platform’s protected storage. Nothing secret is stored on the server. A breach of the server leaks public keys, which are useless to an attacker.

Login. The site sends a random challenge. Her device asks her to unlock it with a fingerprint, face or PIN. That local check only releases the private key to sign the challenge. The signature goes back, and the site verifies it with the stored public key. The biometric never leaves the device.

Back to the fake page

Now replay the month-end email. Farida clicks and reaches the lookalike site. It asks her to sign in. Her browser, not Farida, tells the authenticator which website is asking, and that website name is included in what gets signed. The passkey was created for the genuine domain. The lookalike domain is different, so the browser finds no matching credential and offers nothing. Farida sees no passkey option, and the attacker receives nothing usable.

That property is called origin binding, and it is why people call FIDO credentials phishing-resistant. The signature is only valid for the real site, so a relay attack does not work. The scenario that defeated password-plus-code simply has nothing to capture.

The same design explains two other benefits. Because each site gets its own key pair, there is no credential to reuse, so credential stuffing across sites stops mattering. And because the private key signs a fresh random challenge each time, a recorded login is worthless later.

Where the story gets less tidy

Suppose Farida loses her phone at a railway station. This is the part marketing slides skip.

  • Synced passkeys (kept in a platform account such as an Apple, Google or password-manager vault) reappear on her new device after she signs into that account. Convenient, but now the security of the vault account matters a great deal.
  • Device-bound credentials, like a hardware security key, do not sync. They are stronger in some threat models and a real nuisance if she owns only one. Organisations normally ask for two registered keys.

And then the recovery path. If a site lets Farida bypass the passkey by answering a helpdesk call, or by clicking an email link and typing an SMS code, an attacker will aim at that weaker door instead. A phishing-resistant login is only as strong as the weakest permitted fallback. Many firms therefore require identity checks, manager approval or in-person verification for recovery, and remove SMS codes from the options once passkeys are widespread.

Other limits worth knowing. Passkeys protect the login moment, not what comes after: if malware on Farida’s laptop steals her active session token, the attacker rides in without any login at all. Older applications, command-line tools and some VPN clients may not support WebAuthn yet, so organisations often run passkeys for web sign-in alongside other controls. And “attestation”, the ability to check exactly which authenticator model was used, is a feature for managed fleets, not something consumer sites usually need.

Reading the vocabulary

When you meet the terms, map them like this. The relying party is the website or service. The authenticator is whatever holds the private key. User verification is the local fingerprint, face or PIN step. Discoverable credentials are those that let a person sign in without even typing a username. Behind all of them sits the same simple idea: prove possession of a private key to the right website, and never hand over a secret that someone could be tricked into typing elsewhere.

Farida, by the way, ends her month-end with the same tired eyes. But the fake page gave the attacker nothing, because there was nothing for her to give.

Figures and names in the examples are illustrative.

Leave a Reply

Trending