Phishing-resistant multi-factor authentication (MFA) is a form of sign-in that cannot be tricked into working on a fake website. It uses cryptography that ties the login to the real site’s address, so even if a user is fooled into trying to sign in on a lookalike page, nothing the page captures can be reused on the genuine one.
This is different from the MFA most people know. Text-message codes, one-time passwords from an authenticator app, and push-notification approvals all add a second step, but a convincing fake page can collect and relay them in real time. Phishing-resistant methods remove that weakness. The leading examples are FIDO2 security keys, passkeys, and certificate-based authentication such as smart cards.
This page explains how the common MFA types differ, how phishing-resistant authentication works, what it does not protect against, and how to deploy it without locking people out.
Why ordinary MFA can be phished
MFA is built on the idea that an attacker who steals a password still lacks a second proof. That holds against simple password theft. It breaks against adversary-in-the-middle phishing, in which a fake sign-in page sits between the victim and the real service:
- The victim clicks a link and lands on a convincing copy of the login page.
- They type their username and password. The fake page forwards these to the real site.
- The real site asks for the second factor. The fake page shows the same prompt.
- The victim enters the code, or approves the push. The fake page forwards it.
- The real site signs in and issues a session. The attacker’s server captures it. See session hijacking.
The user did everything “right” by their own understanding. The weakness is that a code or an approval has no connection to where it is being entered.
A related technique, push fatigue (also called MFA bombing), floods a user with approval requests until they tap “approve” to make them stop.
How phishing-resistant authentication works
Phishing-resistant methods use public-key cryptography instead of a shared code.
- When you register a security key or passkey with a service, your device generates a key pair. The private key stays on the device and never leaves it. The public key is stored by the service.
- When you sign in, the service sends a challenge. Your device signs it with the private key, after you confirm with a touch or biometric.
- Crucially, the signature is bound to the origin, the exact web address of the site that asked. A fake site at a different address receives a signature that is valid only for that wrong address. The real site rejects it.
Because the secret is never typed, there is nothing to read aloud, nothing to paste into a fake page, and nothing to intercept. The standards behind this are WebAuthn (a web standard for browsers) and CTAP (the protocol between the browser and the key), together branded as FIDO2. See the FIDO Alliance’s overview of passkeys.
Types of MFA compared
| Method | Phishing-resistant? | Notes |
|---|---|---|
| SMS or voice code | No | Can be relayed to a fake page, and attackers can also take over phone numbers by SIM swapping. |
| Email code | No | Relayable, and only as strong as the mailbox. |
| Authenticator app one-time code (TOTP) | No | Better than SMS, still relayable. |
| Push approval | No | Vulnerable to fatigue and to relay. Number matching helps but does not remove the issue. |
| FIDO2 security key (hardware) | Yes | Private key held in a dedicated device. Strongest against remote attack. |
| Passkey (device-bound or synced) | Yes | Uses the same standard. Synced passkeys are convenient, and depend on the security of the sync account. |
| Smart card or certificate (PKI) | Yes | Common in government and large enterprises. |
NIST’s digital identity guidance, SP 800-63B, describes authenticator assurance levels and notes which authenticators resist phishing. In the United States, CISA’s fact sheet on implementing phishing-resistant MFA gives the same ranking and recommends FIDO or PKI-based approaches for the highest-risk users.
What phishing-resistant MFA does not stop
Being clear about the limits is part of using it well.
- Malware on the device. An infostealer that copies a session after you have signed in does not need to defeat your key. See session hijacking. Use endpoint protection and short-lived sessions.
- A stolen, unlocked device or a coerced user. The key proves possession and presence, not intent.
- Weak recovery. If an attacker can reset your account by calling the help desk or using a phone-number recovery path, they bypass the key. Recovery must be as strong as sign-in.
- Excess access after sign-in. MFA proves who is signing in. It says nothing about what that identity can reach. A phished administrator who passed a strong check still holds administrator rights. This is where least privilege and zero standing privilege apply.
- Applications that do not support it. Legacy systems that accept passwords alone remain a weak point.
Think of phishing-resistant MFA as the strongest lock on the front door. It lowers the chance that a stolen login works. It does not change what is inside the house.
How to roll it out
Step 1: Prioritize by risk
Start with the people and systems where a stolen login does the most damage:
- Administrators of your identity provider, cloud accounts and source control
- Engineers with production access
- Finance, executive and HR users, who are common phishing targets
- Everyone else
Step 2: Check what your identity provider supports
Most major identity providers support FIDO2 and passkeys, and many let you require them through a policy, often called an authentication strength or sign-in policy. Confirm which applications behind your identity provider can enforce them, and which accept weaker fallbacks.
Step 3: Remove the fallbacks
A user with a security key who can also sign in with an SMS code is only as strong as the SMS code. For high-risk groups, require phishing-resistant methods and disable weaker ones. Before you do, check that people have a backup.
Step 4: Issue backups and plan recovery
Give each user at least two authenticators, for example a hardware key plus a second key or a platform passkey. Write the recovery process down. It should require identity verification that an attacker cannot easily fake, and it should be logged and approved.
Step 5: Add device and session controls
Combine strong sign-in with conditional access that requires a managed, compliant device, and with shorter session lifetimes for sensitive applications.
Step 6: Cover non-human access separately
Service accounts, pipelines and AI agents do not use passkeys. They need short-lived, scoped credentials. See non-human identities.
Step 7: Measure
Track the share of privileged users on phishing-resistant methods, the number of fallback sign-ins, and the volume of MFA-related help desk requests.
How Cloudanix helps
Cloudanix does not issue or validate authenticators. That is the job of your identity provider. It works with the providers you already use, including AWS IAM Identity Center, Entra ID, Okta, JumpCloud, Google Cloud Identity, OneLogin and Ping, and it addresses what strong sign-in does not.
- Limiting what a successful sign-in can do. Cloud JIT removes standing administrator access to AWS, Azure and GCP. Even a user who signed in with a passkey holds no elevated access until an approved, time-boxed request is active. Database JIT, VM JIT and Kubernetes JIT do the same for databases, servers and clusters.
- Same sign-in experience. Because JIT works through your identity provider and existing single sign-on, engineers keep their tools and their strong authentication. Requests can be made from Slack or the console.
- Machine identities. Agentic JIT and Coding Agent JIT give pipelines and AI agents credentials that last for a task, covering the identities that cannot use passkeys.
- A record with the real person. Every request, approval, session and revocation is logged with the human identity to storage you own.
See how a single stolen login plays out in Operation Endgame and 6.4 million stolen logins.
Further reading
- FIDO Alliance: Passkeys
- NIST SP 800-63B: Authentication and Lifecycle Management
- CISA: Implementing Phishing-Resistant MFA (fact sheet)
Frequently asked questions
What is phishing-resistant MFA?
It is a sign-in method that uses public-key cryptography bound to the real website’s address, so a fake page cannot capture anything usable. Examples are FIDO2 security keys, passkeys and smart cards.
Are passkeys phishing-resistant?
Yes. Passkeys use the same FIDO standards as hardware keys, and the signature is tied to the legitimate site. Synced passkeys depend on the security of the account that syncs them.
Is an authenticator app code phishing-resistant?
No. A time-based code can be typed into a fake page and relayed to the real one. It is better than SMS, but it does not resist adversary-in-the-middle phishing.
Does phishing-resistant MFA stop session hijacking?
Not by itself. It protects the sign-in. If malware steals the session afterward, the attacker reuses it. Combine it with endpoint protection, short sessions and device controls.
Is MFA enough to protect cloud administrators?
It is necessary and not sufficient. Also remove standing administrator rights, so a successful sign-in alone does not grant powerful access. See just-in-time access.