Skip to content
All posts

Security Awareness

Helpdesk Vishing: Training the People Attackers Call

Attackers phone the helpdesk to reset passwords and MFA. How these calls work, where verification fails, and how to train staff to hold the line.

AAArslan AminChief Operating Officer

· 10 min read

On this page

An attacker doesn't need to beat your multi-factor authentication if your helpdesk will reset it for them. One phone call that sounds like a stressed employee locked out before a deadline can be enough to get a new password, a new MFA device and full access to the account.

This post is for anyone who runs or funds an IT helpdesk, whether it is in-house, outsourced or one person who "does IT". It explains why helpdesk staff are now a main target, what these calls tend to look like, and how to build verification and training that work at three in the afternoon with a queue of tickets waiting.

Why attackers call the helpdesk

Most organisations have made logins much harder to steal in recent years. Each of those controls shares one weak point: the process for when a real user loses access. Phones get lost, people change devices, and authenticator apps get wiped. Someone has to be able to put things right, and that someone is the helpdesk.

Attackers have noticed. CISA's joint advisory on the Scattered Spider group (AA23-320A) describes callers who first learn what a company's password reset process involves, then phone the IT helpdesk to "reset passwords and/or transfer MFA tokens". The advisory was updated in July 2025 to describe callers posing as employees to have MFA moved to "a device they control". Okta published a similar warning in 2023: callers asked service desk staff to reset all MFA factors on highly privileged accounts, then used those accounts to weaken sign-in rules for everyone else.

MITRE ATT&CK tracks the techniques as Spearphishing Voice (T1566.004) and Impersonation (T1656). The pattern is the same: target the team whose job is to help people who can't prove who they are in the usual way.

What a typical call looks like

The calls we use in awareness exercises, and those in public advisories, share a shape. Knowing it is half the training.

The groundwork

Before calling, the attacker collects what an ordinary helpdesk check asks for. Names, job titles and reporting lines come from LinkedIn and the company website. Employee IDs, office locations and internal jargon can turn up in old job adverts, conference slides and leaked data. If your helpdesk confirms identity with date of birth, start date or manager's name, assume the caller already has them.

The pretext

The story is usually dull and believable. "I've got a new phone and the authenticator didn't carry over." "I'm at the airport and need to get into email before a board call." "My manager said to call you, I'm presenting in ten minutes." The best pretexts match something real, such as a travel week, a known system migration or month-end close.

The pressure

Urgency, seniority and helpfulness do most of the work. The caller may claim to be an executive, or to be calling for one. They may be friendly and apologetic, or short and irritated. Both are designed to make the agent want the call to end well and quickly.

The ask

The request is almost always one of three: reset the password, remove or replace the MFA method, or add a new device. Sometimes it is smaller: "can you add my personal email for now?"

Where helpdesk verification usually fails

When we review helpdesk processes, the gaps are rarely about staff being careless. They come from the process asking people to make judgement calls with weak evidence. The common ones:

Weak controlWhy it failsStronger alternative
Knowledge questions (date of birth, employee ID, manager's name)Most of this is public or easy to findProof from something the person holds, such as a code pushed to an existing enrolled device
Caller ID or "they called from their mobile"Numbers can be spoofed, and SIM swaps move a number to the attackerAgent calls back on a number from the HR system, never one the caller gives
Reset by SMS to the number on fileThe number itself may be the attacker's after a SIM swapVideo check against the ID photo, or an in-person check for sensitive accounts
Manager approval by email or chatThe manager's account may also be compromised, or the request can be fakedManager confirms through a separate channel the agent starts
Same process for every accountAn admin reset carries far more risk than an intern'sStricter checks and a second approver for privileged accounts
Agents allowed to "make an exception"Pressure always finds the exceptionAgents have a scripted way to refuse and escalate, with management backing

Designing a verification process that holds

The aim is a process an agent can follow under pressure without having to judge whether the caller "sounds right". Four principles do most of the work.

1. Prove identity with something the person has, not something they know

The best check uses an authenticator the user already has: a push to the existing enrolled device, a code from the current authenticator app, or a sign-in to a self-service portal that already requires MFA. If the user has lost every factor, that is when stronger checks start, not when the helpdesk improvises. NIST's digital identity guidance (SP 800-63B) treats account recovery as part of authentication, and expects it to be as strong as the factors it replaces.

2. Never trust the inbound channel

The agent should call back on a number taken from the HR system or directory, not one given during the call. For staff who have lost their phone, a video call where the agent compares the face with the ID photo on file, or an in-person visit, is slower but much harder to fake. Okta's guidance after the 2023 attacks recommends this kind of visual check for service desk verification.

3. Tier by risk

Not every reset needs the same friction. A standard user's password reset with the existing MFA still in place carries little risk. Removing all MFA from a global administrator is very different. Write the tiers down:

  • Standard users, MFA intact: self-service or a quick check against the enrolled device.
  • Standard users, all MFA lost: call-back plus a manager confirmation the agent starts.
  • Privileged, finance and executive accounts: video or in-person check, a second approver, and an alert to the security team for every reset.

4. Make "no" safe to say

The single biggest change we see in helpdesk teams is when management says, in writing, that refusing a request is never wrong if the check fails. Give agents a script for the refusal: "I can't complete that on this call. I'll raise a ticket and call you back on the number we hold." Then back them when an executive complains. If the only person who can approve an exception is a senior manager, and that manager gets a call log, most fake urgency goes away.

Training the people attackers call

Generic phishing training is designed for people who receive emails. Helpdesk staff need something different, because their job is to help callers who can't log in. Training that only says "be suspicious" conflicts with what they are measured on.

In our security awareness work with service desk teams, three things make the difference.

Train on the process, not on spotting liars. People are poor at detecting deception by voice. The goal is not to make agents better at judging tone, but to make the verification steps automatic, so the decision never rests on whether someone sounds genuine.

Use realistic calls, with consent. Simulated vishing calls, agreed in advance with management and HR, show how the process holds up in practice. We script the calls around the pretexts above, vary the pressure, and record the outcome against the written procedure, not against the individual. The output is a list of process gaps, such as "the call-back step was skipped when the caller claimed to be travelling", not a list of names.

Debrief quickly and without blame. An agent who was talked into a reset during an exercise has shown you a weakness in the process. If they are punished, the next one will hide it. When we run these exercises, the debrief happens the same day, focuses on which step was missed and why, and ends with an agreed change to the script or the tooling.

A short quarterly refresher keeps the process fresh, and new helpdesk starters should do the verification training before their first call.

Backing the process with controls

People and process come first, but some technical controls make a successful call far less useful to an attacker:

  • Phishing-resistant MFA such as FIDO2 security keys or passkeys for administrators and high-risk roles. CISA recommends these specifically because they resist push bombing and SIM swaps.
  • Alerts on MFA changes. Every new authentication method, especially on a privileged account, should notify the security team and the user through a channel the attacker does not control. This is where helpdesk work meets detection and response.
  • Limited helpdesk permissions. Agents shouldn't be able to reset MFA on global administrators at all. Route those to a separate, smaller team.
  • Waiting periods for sensitive changes. A short delay before a new MFA device can approve high-risk actions gives the real user time to notice.

If an attacker does get through, the account they take over will usually look like a normal sign-in with valid MFA. Our post on detecting stolen-session sign-ins in Entra ID covers what that looks like in the logs. The basics in ten security basics that stop most attacks limit what a single compromised account can reach.

Questions to ask about your own helpdesk

Ask whoever runs your helpdesk, in-house or outsourced. Vague answers mean the process is improvised on each call.

  • Is the identity check for password and MFA resets written down, and does it avoid facts that can be found online?
  • Does the agent ever rely on the number the caller is calling from, or on a number the caller gives?
  • Is there a stricter process for administrators, finance staff and executives?
  • Does every MFA reset on a privileged account alert someone outside the helpdesk?
  • Can an agent refuse a senior person's request without fear of consequences, and is that written down?
  • If the helpdesk is outsourced, does the contract set out the verification steps, and have you seen them followed?
  • When did you last test the process with a realistic, approved call, and what changed afterwards?

The helpdesk exists to get people back to work, and that should not change. What should change is that getting someone back to work always involves proof, not persuasion.

  • Social Engineering
  • Vishing
  • Helpdesk
  • Training
AA

Written by

Arslan Amin

Chief Operating Officer, FutaqSec

Keep reading

Not sure where to start?

A 30-minute scoping call is enough to understand your environment and recommend the right first step. No obligation.

Book a free consultation