Explore our new Cybersecurity Terms Glossary, explained by our experts.

View the Glossary View the Glossary
PRIVIA

Cybersecurity Terms Glossary

What Is an MFA Fatigue Attack?

An MFA fatigue attack floods a user with authentication prompts until they approve one to make it stop. The attacker already has the password; the push is the last barrier.

Illustration for the cybersecurity term MFA Fatigue Attack
Home / Blog / Cybersecurity Terms / What Is an MFA Fatigue Attack?
17 November 2026

What Is an MFA Fatigue Attack?

What Is an MFA Fatigue Attack?

At around eleven at night a contractor’s phone starts buzzing. An approval request from the corporate VPN. Decline. Another. Decline. They keep arriving, one after another, for the better part of an hour. Then a message arrives on WhatsApp from someone claiming to be IT support, explaining that the alerts are a known glitch and will stop once a request is approved. The contractor approves one.

That is how Uber was breached in September 2022, and the pattern has a name: MFA fatigue, push bombing, or prompt bombing. Note what the attacker already had before any of it started—a valid password, bought from an infostealer log or a dark web listing. The second factor was the only thing in the way.

Why It Works

Push approval was designed to remove friction. A prompt appears, the user taps approve, they are in. That design assumes the person seeing the prompt caused it, and under a barrage of repeated requests at an inconvenient hour, the assumption breaks down for entirely human reasons.

The prompt usually carries almost no context—no application name a user would recognise, no location, no indication of what is being authorised. Faced with a notification that says nothing useful and will not stop, people approve to end the interruption, or because they genuinely assume a system is misbehaving. Many organisations have also trained their staff into approving prompts reflexively by generating them several times a day.

The attack costs nothing. Scripts automate it, and the attacker simply needs one approval out of hundreds of attempts.

The Cases Worth Knowing

In every case the social engineering call is what converts persistence into success. Pure prompt flooding often fails; flooding plus a plausible explanation rarely does.

The Controls That Actually Stop It

Number matching. The sign-in screen displays a two-digit number which the user must type into their authenticator app. Approval is no longer a single tap, and a user who is not looking at a sign-in screen has nothing to type. Microsoft made this the default for Microsoft Authenticator in May 2023, and equivalents exist in Duo, Okta Verify, and most other push providers. If you have not confirmed yours is on, that is the first thing to check.

Context in the prompt. Showing the application, the approximate location, and the requesting IP address gives the user something to judge. A prompt that says “sign-in from Lagos” to someone in Leeds is a declined prompt.

Rate limiting and lockout. Cap the number of prompts per account per period and alert when the cap is hit. Repeated denials are one of the highest-signal detections available, because a genuine user rarely declines three prompts in a row.

Phishing-resistant factors. FIDO2 security keys and passkeys remove the attack entirely, because there is no approval to grant—the authenticator signs for a specific origin or does nothing at all. For privileged roles this should be mandatory rather than encouraged.

Risk-based policy. Conditional access that requires a compliant device for sensitive applications means a correct approval from an unmanaged machine still fails.

Telling People the Right Thing

Awareness training helps here more than it does for most threats, provided the message is specific. Two rules cover it. An unexpected prompt is a report, not a decline—declining stops this attempt and tells nobody that the attacker already has the password, which they do. And IT will never contact you to ask you to approve a prompt; anyone doing so is the attacker, whatever number they are calling from.

Make reporting genuinely easy, with a single button or short code, and respond visibly when someone uses it. The organisations that detect this attack early are the ones where the first flooded user reported it within minutes, and that only happens if reporting is faster than ignoring.

The underlying lesson is that a password was already compromised before any prompt appeared. Credential monitoring, blocking the password spraying and stuffing that produces those credentials, and moving away from passwords altogether all reduce the number of times this scenario can begin.

At Privia Security, we run controlled push bombing and voice phishing exercises as part of our social engineering services.

FAQ – Frequently Asked Questions

What Is Number Matching and Does It Work?

Instead of a simple approve button, the sign-in screen shows a two-digit number that the user must enter in their authenticator app. Someone receiving an unexpected prompt has no number to type, so approval becomes impossible rather than merely ill-advised. It is highly effective against push bombing and is enabled by default in most major authenticator applications.

Should Users Decline Unexpected Prompts or Report Them?

Both, and reporting is the more important half. A decline stops that attempt but leaves the attacker holding a valid password and free to try again later. A report tells the security team that a credential is compromised so it can be reset and the account monitored. Make reporting a single tap, and acknowledge it when someone does.

Is SMS a Safer Second Factor Than Push?

No. SMS avoids push bombing but introduces SIM swapping, interception, and relay through phishing pages, all of which are more widely exploited. The right direction is not sideways to SMS but upwards to phishing-resistant authentication with FIDO2 keys or passkeys, with number-matched push as an interim measure where hardware is not yet deployed.

How Do We Detect an MFA Fatigue Attack in Progress?

Alert on repeated denied authentication attempts for a single account in a short window, which is rare in normal use and therefore high-signal. Combine it with sign-in attempts from unfamiliar locations against a correct password, and with user reports. Rate limiting the number of prompts per account per hour both detects and contains the attack.

Related Terms

You May Be Interested In These

What Is Virtualisation Security?

Cybersecurity Terms

17 November 2026

What Is Virtualisation Security?

Read More
What Is a Silver Ticket Attack?

Cybersecurity Terms

17 November 2026

What Is a Silver Ticket Attack?

Read More
What Is Model Watermarking?

Cybersecurity Terms

17 November 2026

What Is Model Watermarking?

Read More