Case file
The helpdesk call that wasn't
The helpdesk analyst had handled thousands of lockout calls, and this one arrived dressed exactly like authority: a senior executive’s name in the caller ID lookup, an executive assistant vouching on the line, and a story — new phone, old authenticator gone, board meeting in forty minutes. The only thing the caller had not counted on was an analyst who followed the script instead of the mood.
What the Sentinel pieced together afterward, working the response engagement that followed, is a case file worth keeping — because nothing in it required a zero-day, and everything in it exploited a process gap.
The anatomy of the call
The attacker already had the password. It had leaked, as passwords do, in an infostealer haul from an unrelated third-party breach, and it still worked — MFA was the only lock left on the door. So the play was the one now documented across the industry’s advisory stream: trigger push notifications repeatedly and let human patience do the brute-forcing. The executive’s phone lit up, again and again, while the “assistant” kept the helpdesk on the line. Approve one prompt to make it stop.
Behind that, the real objective: not the push fatigue itself but the helpdesk. The call was a rehearsal for the recovery path — convincing the service desk that this voice belonged to the name in the directory, so that a fresh authenticator could be enrolled onto the account and the second factor would belong to the attacker permanently. It is the same playbook public reporting has attributed to the groups that breached major companies through helpdesk social engineering in 2022 and 2023 — in one widely reported case, an MFA-fatigue approval against a contractor’s account opened a Fortune 500.1
The ten seconds that mattered
The analyst did three unremarkable things, and they were everything:
- Refused inbound trust. The caller’s number matched nothing on file, and the callback went to the number in the HR system — not the one on the call. The “executive” never came to that phone. The engagement ended there.
- Required number matching. The push prompts carried a two-digit code the real user must type into the authenticator request — a control that turns “approve anything to stop the noise” into “approve the correct number, which the attacker doesn’t know.”2
- Escalated the enrollment request instead of fulfilling it. New device + executive identity = manager-in-the-loop, always, no matter who is asking or how senior the voice sounds.
The MFA prompts kept firing for another hour — pure harassment by then — and no prompt was ever approved.
What the Sentinel rebuilt afterward
The incident closed in an afternoon. The program changes took a quarter, and they are the transferable part of this file:
- The helpdesk is an identity checkpoint, not a courtesy line. Verification scripts with challenge-and-response steps the caller cannot have learned from a leaked org chart, and a hard rule: the helpdesk never initiates credential changes from an inbound request alone. It calls back, on the number of record, every time.
- Number matching everywhere, as an enforced tenant setting rather than a policy document — the difference between the control existing and the control applying.2
- Phishing-resistant authenticators for the roles attackers call about. Executives, finance, IT staff: hardware security keys with provenance binding, at the assurance level the NIST digital identity guidelines describe as resistant to compositional attacks — the class this call belonged to.3
- Helpdesk-specific social-engineering drills, run like the phishing simulations the rest of the staff gets. The analyst who saved this incident had been drilled. The two who would have approved had not.
The transferable sentence
Every identity has a recovery path, and the recovery path is staffed by humans who want to be helpful. An MFA rollout is only as strong as the verification conversation a stranger can have with the helpdesk — which means the helpdesk script deserves the same review rigor as any technical control. In this composite engagement, one page of procedure beat a determined, patient, well-briefed attacker.
The Sentinel’s recommendation has one line: drill the humans who hold the reset button.
This is a labeled composite: details from multiple unrelated engagements are merged and altered per the published methodology. No narrative above describes any single real person, call, or company; the techniques and controls are real and sourced below.
Sources & attributions
- Public reporting on the September 2022 Uber intrusion via contractor credentials and MFA-fatigue approval (e.g., Bloomberg’s original reporting) and the 2023 FBI/CISA advisories documenting helpdesk social-engineering techniques attributed to “Scattered Spider” — cisa.gov/news-events.
- Microsoft Entra documentation, Number matching in push notifications — enforced tenant setting that defeats push-bombing approval — learn.microsoft.com.
- NIST SP 800-63B, Digital Identity Guidelines: Authentication and Lifecycle Management — authenticator assurance levels and phishing-resistant authenticator classes — pages.nist.gov/800-63-3/sp800-63b.html.
Footnotes
-
Bloomberg reported the 2022 Uber breach began with MFA push-fatigue against a contractor’s account; the 2023 advisories extend the pattern to helpdesk-driven authenticator enrollment. ↩
-
Number matching requires the user to enter the code shown on their sign-in screen into the authenticator prompt, so an unsolicited flood cannot be silenced by blind approval. ↩ ↩2
-
Composite case file: see the inline notice and Disclosures — Anonymization. ↩