Phishing-resistant MFA (what NIST SP 800-63B counts)
Not every second factor stops phishing. If the user types or relays a code, a fake site can pass it on. Phishing resistance comes from cryptography bound to the real site or session.
Most MFA rollouts stop at 'a second factor is on'. A phishing page can sit between the user and the real login page and pass along whatever the user enters, code included. The useful question is not whether MFA is enabled, but whether it would still hold if the user were on a convincing fake site.
What NIST SP 800-63B says
NIST SP 800-63B defines phishing resistance as the ability of the authentication protocol to stop authentication secrets and valid authenticator outputs being disclosed to an impostor verifier, without relying on the user noticing. How the user got to the fake site does not matter: a search result and an email link are both phishing.
NIST says phishing resistance requires single- or multi-factor cryptographic authentication. Authenticators where the user manually enters an output, such as out-of-band and one-time-password (OTP) authenticators, are not phishing-resistant, because typing a code does not bind it to the session being authenticated; an impostor can relay it to the real site.
NIST recognises two methods. Channel binding ties the authenticator output to the specific protected channel, for example client-authenticated TLS as used by PIV and CAC smart cards. Verifier name binding ties the output to the verified identity of the site, which is how WebAuthn/FIDO2 authenticators work. NIST rates channel binding as more secure because it is not exposed to misissued or misused verifier certificates, but both meet the requirement.
Levels matter. At AAL2, NIST requires applications to offer a phishing-resistant option. At AAL3, a phishing-resistant authenticator with a non-exportable key is required.
Syncable passkeys (keys copied to other devices through a cloud sync fabric) can support AAL2 under NIST's conditions: keys stored in the sync fabric must be encrypted, access to them must be protected by AAL2-equivalent MFA, and private-key operations must happen on the local device. Because syncing means the key can be exported, synced keys do not meet AAL3.
CyberStack guidance (our own advice, not NIST's)
Start with the accounts an attacker wants most: administrators, email, remote access, cloud consoles and the identity provider itself. Move them to FIDO2 security keys or device-bound passkeys before anyone else.
Close the side doors. If a user can still choose SMS, push or a typed code when the strong factor is 'unavailable', an attacker will choose it for them. Recovery and help-desk reset flows need the same strength as the login.
Measure it. Report the share of privileged and remote-access sign-ins that used a phishing-resistant method, and treat the rest as open risk with an owner and a date.
See also:
Fact source: NIST SP 800-63B-4 — Digital Identity Guidelines: Authentication and Authenticator Management.
