MFA Fatigue Attacks: Push Bombing and MFA Methods Ranked
Skip to content

MFA Fatigue Attacks

An MFA fatigue attack is an attempt to get past multi-factor authentication by sending a user repeated sign-in prompts until one is approved. It is also called push bombing or MFA bombing.

The attacker already has the password. The prompts are the last obstacle, and the attack relies on a tired or confused person tapping approve. Which authentication method the organization uses decides whether the attack can work at all, so this guide ranks the methods.

One Attack, Minute by Minute

The target is a systems administrator at an organization whose sign-in provider sends simple approve-or-deny prompts and does not enforce number matching. The attacker bought the password from a stealer log two days earlier.

1
01:12The attacker signs in with the correct password. The administrator’s phone shows an approval prompt. It is ignored.
2
01:13 to 01:40The attacker repeats the sign-in every minute or two. More than twenty prompts arrive. Each is recorded as a failed second factor.
3
01:41The prompts stop. The attacker switches channel.
4
01:44A message arrives on a chat application, claiming to be from the IT help desk. It says a sync fault is causing the prompts and asks the administrator to approve the next one to clear it.
5
01:46One more prompt arrives. The administrator approves it.
6
01:47A sign-in succeeds from a residential proxy address in the administrator’s own city.
7
01:51A phone number is added to the account as a second sign-in method. From now on the attacker’s own phone can answer the challenge.
8
02:05The attacker signs in to the VPN and begins exploring the network.

From the first prompt to the successful sign-in took 35 minutes and needed no fake website and no technical skill beyond buying a password. The decisive step was the message at 01:44, which gave a reason to approve.

How the Technique Emerged

Push notifications were introduced to make multi-factor authentication easier. Typing a six-digit code was a nuisance, and a single tap on a phone was not.

The convenience created the weakness. A prompt that takes one tap to approve takes one tap to approve by mistake, and nothing in the original design connected the prompt to the person actually signing in. Anyone with the password could make the phone buzz.

The technique came to wide attention in 2022, when several large technology companies disclosed intrusions that began this way. In one, the company said that an attacker had obtained a contractor’s password, sent repeated approval requests, and gained access when the contractor accepted one. The pattern in each case was the same: a purchased or phished password, a stream of prompts, and often a message posing as IT support.

Government advisories now list the method explicitly. The joint advisory on the Scattered Spider group, issued by agencies in the United States, Australia, Canada, and the United Kingdom, names push bombing among the group’s techniques for getting past multi-factor authentication.

The industry response was number matching, which ties the approval to the sign-in screen. It is the reason a simple fatigue attack no longer works against a properly configured Microsoft tenant, and the reason the attack persists wherever older methods remain enabled.

Where the Password Comes From

An MFA fatigue attack is the second half of an intrusion. The first half is obtaining the password, and the response depends on which source it was.

A stealer log

The password was copied from a device the user works on, and that device will leak the next password too until it is rebuilt. The infostealer guide covers the response.

Reuse or spraying

The password was reused from a breached service or guessed in a spray, so other accounts are likely exposed the same way. The password spraying and credential stuffing guide explains both.

Plain phishing

A credential-harvesting page collected the password without relaying the session, and other recipients of the same message need checking.

An earlier incident

Credentials taken in a previous compromise were never reset.

Resetting the one account without asking where the password came from leaves the cause in place.

Why People Approve

Habit

Staff approve genuine prompts many times a day and stop reading them.

Timing

Some attackers send prompts at the start of the working day, when a real sign-in is likely in progress and one more prompt looks routine.

Exhaustion

A phone that buzzes through the night gets a tap to make it stop.

Assumed fault

Repeated prompts look like a technical glitch, and approving seems the way to clear it.

Authority

A caller or message from the help desk supplies a plausible instruction.

Training reduces the risk and cannot remove it. A control that depends on every person refusing every unexpected prompt, at any hour, will fail eventually. The durable fix is a method with nothing to approve.

MFA Methods Ranked

Methods differ in how they hold up against prompt bombing and against the other common ways around a second factor, relay phishing and SIM swapping.

MethodPrompt bombingRelay phishingSIM swapping
Push notification, approve or denyVulnerableVulnerableResistant
Phone call, press a key to approveVulnerableVulnerableVulnerable
Text message codeResistantVulnerableVulnerable
Authenticator app codeResistantVulnerableResistant
Push notification with number matchingResistantVulnerableResistant
Passkey or hardware security keyResistantResistantResistant
Certificate or device-bound sign-inResistantResistantResistant

Codes and number matching resist prompt bombing and still fall to relay phishing, where the user enters the code or number on a page controlled by the attacker. Passkeys, security keys, and certificate or device-bound sign-in are the only methods that resist every attack in the table.

In Microsoft Entra ID, a Conditional Access authentication strength turns this ranking into policy. The built-in phishing-resistant strength accepts only passkeys, Windows Hello for Business, and certificate-based authentication for the applications and roles it covers.

Number Matching in Practice

Number matching changes what the user is asked to do. The sign-in screen displays a two-digit number, and the phone asks the user to type that number to approve. Someone who is not looking at a sign-in screen has nothing to type.

Microsoft has applied number matching to Microsoft Authenticator push notifications for all users since May 2023. Organizations using other providers, or other methods, should confirm that an equivalent is switched on.

It leaves gaps worth knowing about.

The attacker can supply the number

A caller posing as IT support can read the number out and ask the user to enter it. This requires a live conversation and a convincing story, which raises the cost of the attack a great deal.

Relay phishing is unaffected

On a proxy phishing page, the number shown to the user is the real one, relayed from the real service. The user types it, and the session goes to the attacker, as the session hijacking guide describes.

Additional context narrows the first gap. When the prompt shows which application is being signed in to and where the request comes from, a user asked to approve a VPN connection from another continent has a clear reason to refuse.

Moving to Passkeys and Security Keys

Passkeys and hardware security keys work differently from every other method in the table. There is no code to type and no prompt to approve. The user confirms with a fingerprint, face, or device PIN, and the device proves its identity to the service using a key that never leaves it.

The key is bound to the genuine website, so it will not answer a lookalike, and there is no shared secret that a user could be talked into revealing.

A rollout usually runs in this order.

1
Administrators firstThey are a small group, the highest risk, and the most likely to be targeted by name.
2
High-risk roles nextFinance, payroll, executives, and their assistants are the accounts most often targeted for fraud.
3
Protect enrollmentRegistering a new passkey should require a managed device or a recent strong sign-in. Otherwise an attacker with a stolen session enrolls their own.
4
Secure recoveryDecide what happens when someone loses their device. If the answer is a phone call to the help desk, verification there has to be as strong as the passkey.
5
Retire the weak methodsRemove simple push approval, phone calls, and text codes for users who have a stronger method. A fallback that remains available is the one attackers will use.
6
Enforce it in policyOffering a passkey is different from requiring one. An access policy has to demand the phishing-resistant method for the covered roles, or the password and prompt remain a way in.

In Microsoft Entra ID, a Temporary Access Pass covers the first and the lost-device cases. It is a time-limited passcode issued by an administrator that lets a user register a passkey without falling back to a weaker method.

What the Logs Show

A fatigue attack is unusually clear in sign-in data, because every prompt is a recorded event.

Correct password, second factor failed, repeatedly

A run of sign-ins for one account shows the first factor succeeding and the second denied or timed out. In the example, there were more than twenty within half an hour.

A source the user never uses

The attempts come from one address or a small set, typically a hosting provider, an anonymizing network, or a residential proxy.

An unusual pattern for the user

The burst falls outside the user’s normal sign-in times, or arrives as a cluster no real sign-in would produce.

Failure, then success

A single approved prompt at the end of a run of failures is the signature of a successful attack.

A new method minutes later

Security information is registered shortly after that success.

A user report

Where the reporting option is enabled, a user who marks a prompt as suspicious creates a record and raises their risk level.

In Microsoft Entra ID the failed prompts appear as sign-in failures with error code 500121, authentication failed during the strong authentication request. The authentication details on each record say whether the user denied the prompt or did not respond.

Every unexpected prompt means the password is known. The account needs a password reset whether or not a prompt was approved.

Investigating After the Fact

Many fatigue attacks are found days later, during another investigation. Once the run of failed prompts has been found in the sign-in log, the remaining questions are about people and scope.

Look for the conversation

Check chat, email, text, and phone records around the time of the approval for a message claiming to be from IT. Ask the user directly, because most remember.

Check the help desk

Look for a call or ticket about the account in the same period. An attacker who failed with prompts often tries the help desk next.

Trace the password

Check the user’s devices for malware, breach data for the address, and the sign-in logs for an earlier spray.

Look at the neighbors

The same source working through other accounts with correct passwords means the attacker holds a list of credentials.

The findings decide the scope. One user approving one prompt is an account takeover. The same source holding correct passwords for ten accounts is a credential exposure across the organization.

Settings That Close the Gap

Beyond the choice of method, a few settings in the identity platform shorten the attack or stop it before a prompt is sent.

Report Suspicious Activity

Users can be given a way to report a prompt they did not start. In Microsoft Entra ID the report marks the user as high risk, which triggers whatever policy applies to risky users, such as blocking sign-in until the password is changed.

System-Preferred Method

Where a user has several methods registered, the sign-in may default to an older and weaker one. The system-preferred setting in Microsoft Entra ID prompts with the strongest method the user has registered. A weaker method that stays registered can still be chosen as an alternative, which is why the weak methods have to be retired as well.

Limit the Attempts

Risk-based access policies can block sign-ins from unfamiliar or anonymous sources before any prompt is sent. Where the provider supports it, lockout after repeated failed second-factor attempts stops the stream of prompts.

Require a Managed Device

Where sign-in requires a compliant device, a correct password and an approved prompt are still not enough from the attacker’s machine.

Alert on the Run

A run of failed second factors on one account, and any success that ends such a run, should raise an alert on its own. Waiting for the user to report leaves a night-time attack unanswered until morning.

When a Prompt Was Approved

An approval from a source the user does not recognize is an account takeover. Act in this order.

1
Block sign-in if the attacker is active or the account is privilegedIn the example the account belongs to an administrator, so it is blocked first.
2
Reset the passwordThe attacker already had the password, so it has to change before the sessions are cut.
3
Revoke the account’s sessionsThis ends the session opened by the approved prompt.
4
Remove what the attacker addedRecord each item, then delete any authentication method, device, rule, or application grant the user does not recognize.
5
Confirm in the sign-in logsCheck that nothing from the attacker’s sources succeeds afterward, including on the VPN.
6
Check other accountsThe same source may be working through a list of users with stolen passwords.

The account takeover guide explains what each step removes. The source of the password decides what else has to be fixed.

What to Tell Staff

Technical settings do most of the work, and staff still need to know what to do when something unexpected appears on their phone. The message should be short enough to remember.

A prompt you did not start means someone has your password

Deny it and tell IT, even if it is the middle of the night.

IT will never ask you to approve a prompt, read out a code, or enter a number they give you

Anyone who does is an attacker, whatever name they use.

Repeated prompts are an attack, not a fault

Approving one will not make them stop. Reporting will.

If you approved one by mistake, say so immediately

Minutes matter, and nobody is in trouble for reporting.

Check the details on the prompt

If it shows an application or a location that is not you, deny it.

The message sticks when reporting is easy and safe. Give people one obvious way to report, such as a button in the authenticator app or a single help desk line. Respond visibly when someone reports, so that reporting feels worthwhile. Treat a mistaken approval as useful information, because staff who fear blame stay quiet, and a takeover that could have been contained in ten minutes runs for a day.

Administrators and finance staff should hear this in person, with examples, since an attacker will prepare a convincing story for them.

The Help Desk Is the Next Target

Attackers who cannot get a prompt approved change approach. They call the help desk, pose as the locked-out user, and ask for a password reset or a new authenticator. The same information that let them write a convincing message at 01:44 lets them answer basic identity questions.

Help desk procedures need the same strength as the sign-in they can override. Requests to reset credentials or enroll a new device should require proof an outsider cannot supply, such as a call back to a number on record, a manager’s confirmation, or verification through an existing strong method.

MFA Fatigue Detection With ThreatDefence

ThreatDefence monitors Entra ID sign-ins and changes to authentication methods 24x7. Those are the traces a successful prompt attack leaves: an approval the user did not start, then a new method registered by the attacker. Our SOC contains a confirmed takeover under agreed authority by revoking sessions and removing the method.

A free Microsoft 365 forensic review reports MFA coverage and risky sign-ins across your tenant, including which accounts still rely on phone call or text message methods.

A Microsoft 365 forensic scan in the ThreatDefence portal: six months of activity analyzed, sign-ins by risk, MFA coverage, app consents, findings by severity and the response taken

Experiencing a breach?

Our incident response team answers 24x7, with or without a retainer.

Talk to Our Experts

Ask about a free Microsoft 365 forensic review, covering six months of activity in your tenant.

    I'm an MSP, MSSP, or reseller

    By submitting this form, you accept the terms in our privacy policy.

    Frequently Asked Questions

    What is an MFA fatigue attack?

    An MFA fatigue attack is a run of sign-in attempts by someone who already holds the user’s password, each one sending an approval request to the user’s phone until one is accepted. Push bombing, MFA bombing, and prompt bombing are other names for it. It works only against methods that can be approved without seeing the sign-in screen: simple push notifications and phone call approval.

    What should a user do with an unexpected MFA prompt?

    Deny it and report it to IT or security straight away, by the route the organization has published and never by replying to a message about the prompt. An approval request the user did not start means someone has typed the correct password. The password is reset even when every prompt was denied, and the report lets the security team look for the same source against other accounts.

    Does number matching stop MFA fatigue attacks?

    It stops the simple form, where the user taps approve to make the prompts stop, because approval now requires a number that only the person at the sign-in screen can see. It leaves the attack where a caller reads the number out to the victim, and it does nothing against a phishing page that relays the number. Passkeys and security keys close both.

    Can attackers bypass MFA without the user approving anything?

    Yes. A relay phishing page or infostealer malware can take the session issued after a genuine sign-in, a SIM swap redirects text message codes, and a help desk can be talked into registering a new method. None of these sends the user a prompt to refuse. Session theft is covered in the session hijacking and token theft guide.

    Why do MFA fatigue attacks happen at night?

    A phone that buzzes repeatedly at 1 a.m. is more likely to be silenced with a tap than read, and the person holding it is alone, with no colleague or help desk to ask. An organization without out-of-hours monitoring also gives the attacker until morning before anyone looks at the sign-in log. Other attackers choose the start of the working day, when one more prompt looks routine.

    Which MFA methods are resistant to fatigue attacks?

    Passkeys, hardware security keys, and certificate-based sign-in resist them completely, because there is no request to approve. Authenticator app codes, text message codes, and push with number matching resist simple prompt bombing, since the user has to supply something the attacker cannot see, though each can still be phished through a relay page. Simple push and phone call approval do not resist it.

    How do we know if an MFA fatigue attack succeeded?

    Look for a successful sign-in from the same source at the end of the run of failed prompts. In Microsoft Entra ID the failures carry error code 500121, and the authentication details show whether each prompt was denied or ignored. A success followed by a new authentication method, a VPN connection, or mailbox activity from that source confirms a takeover. A run that ends with no success still means the password is known.