The most damaging intrusions of 2023 to 2025 did not start with malware: they started with a phone call to a help desk agent who reset a password or MFA factor for someone pretending to be an employee. MGM Resorts, Co-op and, according to reporting, Marks and Spencer were all hit through impersonation of users or IT support, and CISA's Scattered Spider advisory describes the pattern in detail. This guide is for service desk managers, security awareness leads and IAM teams. It covers how these calls work, what help desk agents must be trained to spot and refuse, how to train ordinary employees against callers pretending to be IT, how to extend this to outsourced service desks, and how to run authorised vishing drills and measure the results. The design of the verification process itself is covered in the Account Recovery and Help Desk Security Guide; this page is about the people who operate it and the people who are targeted.
Key takeaways
- A reset is a login. If a caller can get a password or MFA factor reset with facts found online, the help desk is the weakest authentication path in the organisation. Train agents to treat every reset as an authentication decision, not a customer service task.
- Two audiences, two scripts. Help desk agents are called by fake employees; employees are called by fake IT. Both need short, specific rules, and the rules must match each other.
- Pressure is the signal. Urgency, seniority, a new starter who cannot reach their manager, a travelling executive with a lost phone: these are the scripts in the advisories. Agents need explicit permission to slow down and refuse.
- Privileged accounts need a different path. Okta and Microsoft both recommend scoping help desk roles so they cannot reset highly privileged administrators. Awareness cannot compensate for agents who hold that power.
- Outsourced desks are in scope. If a third party answers your service desk phone, its agents need your training, your verification rules and your drill programme.
How help desk social engineering works
This section is descriptive, so agents and trainers recognise the pattern. It does not describe how to carry it out.
| Phase | What happens | ATT&CK v20 | What breaks it |
|---|---|---|---|
| Research | The caller gathers names, roles, manager names, employee IDs and personal details from LinkedIn, breach data and earlier phishing. Microsoft observed Octo Tempest mimicking employees' way of speaking. | T1589 Gather Victim Identity Information; T1598.004 Spearphishing Voice | Do not use facts that can be researched (name, ID, manager, date of birth) as verification |
| Pretext call | The caller poses as an employee who has a new phone, is travelling, is a new starter or is a senior executive in a hurry. CISA describes actors persuading help desks to reset passwords or MFA tokens using victims' PII and security answers. | T1566.004 Spearphishing Voice | Verification that cannot be satisfied by voice alone |
| Reset or factor change | The agent resets the password, removes MFA factors or registers a new device. Okta reported attackers pressuring service desk staff to reset all MFA factors for highly privileged users. | T1556.006 Multi-Factor Authentication; T1078 Valid Accounts | Scoped help desk roles, approval for privileged resets, alerts on factor resets |
| Calls to employees | The same groups call or text employees posing as IT, asking them to install a remote access tool, read out a one-time code or approve a push. | T1219 Remote Access Tools; T1621 Multi-Factor Authentication Request Generation | Employee rule: IT will never ask for codes or remote access you did not request |
| Third-party route | The caller targets a contractor or outsourced help desk with weaker processes. | T1199 Trusted Relationship | Same rules, training and monitoring for outsourced desks |
| Phone number takeover | A mobile carrier is persuaded to move the victim's number, defeating SMS codes and callback to that number. | T1451 SIM Card Swap (mobile) | Phishing-resistant MFA; do not rely on SMS for recovery of privileged accounts |
The Scattered Spider Threat Actor Profile covers the group behind many of these cases. For the broader vishing picture, see the Vishing and Smishing Guide.
What help desk agents must be trained to spot
Agents need a short list of red flags tied to an action, not a list of things to "be aware of". Put the action in the ticketing system as a required step.
| Caller says | Red flag | Agent action |
|---|---|---|
| "I have a new phone and lost my authenticator" | The most common pretext for a factor reset | Follow the strong verification path; never re-register a factor on voice verification alone |
| "I am the CFO, I am about to board, I need this now" | Seniority plus urgency | Senior and privileged users go to the high-assurance path; seniority never shortens it |
| "I started this week and cannot reach my manager" | New starter pretext, noted by Microsoft for Octo Tempest | Verify through the manager or HR via a contact already on file, initiated by the agent |
| "Please send the code to this other number or email" | Redirecting recovery to an attacker-controlled channel | Only use contact details already held in the HR or identity record |
| "My manager already approved it, check the ticket" | Approval may come from a compromised manager account, as Microsoft observed | Confirm approval through a separate channel, not a reply in the same ticket or chat |
| Caller is calm but cannot answer an out-of-band check, then offers more personal details | Substituting researched facts for verification | Do not accept extra knowledge facts as a substitute; end the call and log it |
| Caller asks to remove MFA from an admin or service account | Targeting privileged or non-human access | Refuse at tier 1; escalate to the IAM or PAM team |
Agents should be told, in writing and by their manager, that ending a call and asking the person to come back through the strong path is always the correct outcome when they are unsure. Measure agents on following the process, not on first-call resolution for reset tickets.
Verification tiers agents should apply
| Account type | Minimum verification for a password or MFA reset |
|---|---|
| Standard user, password only | Self-service reset with an existing phishing-resistant factor, or agent-initiated callback to the number in the HR record plus a manager confirmation |
| Standard user, MFA factor reset | Live video with a government ID matched to the HR record, or in-person verification, or manager confirmation through an authenticated channel; issue a short-lived, single-use enrolment credential rather than disabling MFA |
| Privileged administrator | In-person or high-assurance video verification plus approval by a second member of the IAM team; reset handled by a separate privileged team, not tier 1 |
| Executive and finance approver | As for privileged administrators, because they are the most researched and impersonated |
| Service accounts and other non-human identities | Never reset on a phone request; changes only through the owning team's change process and secrets manager |
Okta's advisory recommends visual verification for the help desk and recovery flows that start with the strongest available authenticator. For the deepfake risk to video verification, see the Deepfake-Resistant Identity Verification Guide. NIST SP 800-63B-4 (final, July 2025) is the reference for authenticator management and recovery assurance.
Training employees against fake IT calls
CISA describes actors calling or texting employees while posing as IT or help desk staff to collect credentials or get remote access tools installed, and asking employees to read out one-time passcodes. Microsoft observed callers persuading users to remove their FIDO2 token. Give every employee these rules, repeated in onboarding and in each awareness cycle:
- IT will never ask for your code: no genuine support person will ask you to read out a one-time code, approve a push you did not start or tell them your password.
- IT will not ask you to install remote access tools: unless you raised a ticket and the tool is the approved one named on the intranet.
- Hang up and call back: if someone calls claiming to be IT, end the call and contact the service desk through the number on the intranet or the self-service portal.
- Do not remove your security key or authenticator: on anyone's instruction by phone, chat or Teams.
- Report it: report the call or message using the report button or the security contact, even if you did nothing. Okta's system log captures user reports as user.account.report_suspicious_activity_by_enduser where that feature is enabled.
Recent campaigns combine email bombing with a follow-up Teams call from "IT"; see the Email Bombing and Help Desk Impersonation over Teams Guide. For push approval attacks, see the MFA Fatigue and Bypass Guide.
Outsourced and contracted service desks
Many organisations outsource first-line support. M&S told MPs that attackers impersonated one of its third-party users; reporting links this to its outsourced IT help desk, which M&S has not confirmed. Whatever the route, a contracted desk is part of your authentication path.
- Contract: require the provider to follow your verification standard and to train its agents on it. See the Third-Party Security Contract Clauses Template.
- Same tooling: agents should use your identity verification tooling and ticket workflow, so every reset leaves evidence you can audit.
- Scoped roles: provider agents get the narrowest help desk role you have, with no ability to reset privileged accounts.
- Drills: include the provider's agents in your vishing simulations, with the provider's agreement written into the contract.
- Reporting: the provider must report suspected social engineering attempts to your security team within [an agreed time].
Running authorised vishing drills
- Get written authorisation: agree scope, scripts, numbers to be called and the escalation route with HR, legal and the service desk owner. Tell the desk's managers, not the agents.
- Use realistic, ethical pretexts: lost phone, new starter, travelling executive. Do not use pretexts involving bereavement, illness or threats.
- Test the process, not the person: score whether the verification step was followed, whether the call was escalated and whether it was reported. Never use results for discipline on a first failure.
- Debrief quickly: tell agents the same day that a drill took place and walk through what the correct handling would have been.
- Fix the process first: if agents fail because the process allows a voice-only reset, change the process before retraining.
- Repeat: run drills at least [quarterly] for the help desk and include outsourced desks. See the Social Engineering and Physical Penetration Testing Guide for engaging external testers.
Where identity fits
Help desk staff are themselves privileged identities: a tier 1 agent who can reset any password or MFA factor holds one of the most powerful roles in the directory. Microsoft recommends scoping help desk privileges to exclude Global Administrators, using Administrative Units and Privileged Identity Management; Okta recommends constraining help desk roles with resource sets that exclude highly privileged administrators and requiring dual authorisation for just-in-time elevation. Non-human identities should sit entirely outside help desk reset rights: a service account password or API key reset requested by phone is always a red flag. See the OWASP NHI Top 10 NHI10 Human Use of NHI Guide.
Detection and measurement
Log and alert on these events, and correlate them with service desk tickets:
- Full MFA factor resets: Okta user.mfa.factor.reset_all; factor activation and deactivation under system.mfa.factor; the equivalent security information registration and password reset audit events in Microsoft Entra ID.
- New identity providers or federation changes: Okta system.idp.lifecycle.create, which was part of the 2023 cross-tenant impersonation pattern.
- Sign-in after reset from a new device, location or anonymising proxy: Okta user.session.start with securityContext.isProxy true.
- Reset without a matching ticket or verification record: the strongest single signal of a bypassed process.
Metrics that show the programme is working:
- Percentage of MFA resets with a recorded strong verification step (target 100 per cent).
- Vishing drill pass rate for help desk agents, by team and provider, over time.
- Number of social engineering attempts reported by agents and employees per month.
- Number of privileged accounts that tier 1 agents can still reset (target zero).
- Median time from a suspicious reset to security team review.
Framework mapping
| Area | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 | Other |
|---|---|---|---|
| Employee awareness of fake IT calls | PR.AT-01 | AT-2, AT-2(3) | ISO/IEC 27001:2022 A.6.3; ATT&CK M1017 |
| Help desk agent training | PR.AT-02 | AT-3 | ISO/IEC 27001:2022 A.6.3 |
| Caller verification and proofing | PR.AA-02 | IA-12 | NIST SP 800-63B-4 |
| Authenticator reset and recovery | PR.AA-03 | IA-5 | ISO/IEC 27001:2022 A.5.17 |
| Monitoring resets and sessions | DE.CM-03 | AU-6 | D3FEND D3-UBA User Behavior Analysis |
Common failure patterns
- Knowledge-based verification: asking for employee ID, manager name or date of birth, all of which can be researched. CISA notes actors gathered PII and security question answers before calling.
- Measuring speed: agents are scored on call handling time, so slowing down to verify is penalised.
- Tier 1 can reset admins: one successful call becomes administrator access, as in the MGM Resorts breach, where the attackers claimed a help desk call led to Okta and Azure administrator access.
- Disabling MFA instead of re-enrolling: the agent removes MFA "temporarily" so the user can log in, leaving the account open.
- Training employees but not agents: all-staff phishing training never covers the agents who hold reset power.
- Outsourced desk outside the programme: the provider's agents have their own scripts and no drills.
Lessons from incidents
In the Co-op cyber attack, reporting describes attackers posing as IT or help desk staff and having an employee's password reset, then moving to Active Directory; the data of 6.5 million members was taken. In the Marks and Spencer cyberattack, M&S described "social engineering/impersonation of an individual combined with a third party vector" and estimated a £300 million profit impact. MGM estimated a negative impact of about $100 million after its 2023 incident. The common lesson: the help desk sits on the authentication path, and the same verification rules, training and monitoring must cover every desk that can reset a credential.
Practitioner checklist
- Remove knowledge-based questions from help desk verification for MFA and privileged resets.
- Define verification tiers by account type and enforce them as mandatory ticket steps.
- Remove tier 1 ability to reset privileged administrators and non-human identities.
- Train help desk agents on the red flag table and give written permission to refuse and escalate.
- Give every employee the five fake IT call rules in onboarding and each awareness cycle.
- Write verification, training and drill participation into outsourced service desk contracts.
- Run authorised vishing drills against the help desk at least quarterly and fix process gaps first.
- Alert on MFA factor resets and correlate every reset with a ticket and verification record.
Standards and references
- CISA and FBI: Scattered Spider advisory AA23-320A
- Okta Security: Cross-Tenant Impersonation: Prevention and Detection
- Microsoft Threat Intelligence: Octo Tempest crosses boundaries to facilitate extortion, encryption, and destruction
- NIST: SP 800-63B-4 Digital Identity Guidelines: Authentication and Authenticator Management
- NIST: Cybersecurity Framework 2.0
- NIST: SP 800-53 Rev. 5
- MITRE ATT&CK: Spearphishing Voice
- MITRE ATT&CK: Modify Authentication Process: Multi-Factor Authentication
Related NHI Mgmt Group resources: Account Recovery and Help Desk Security Guide · Help Desk Social Engineering Incident Playbook · Security Awareness and Human Risk Management Guide · Scattered Spider Threat Actor Profile