Out-of-band verification is a security control that confirms a sensitive request, such as a payment, a bank detail change or an access reset, through a separate channel the requester did not choose. It works whether or not a deepfake, a cloned voice or a spoofed email is convincing.
Key takeaways
- Out-of-band verification checks a request on a second channel that was set up before the request arrived, such as a known phone number from the directory.
- It does not rely on spotting the fake, which is why it holds up against deepfake video calls and AI voice cloning.
- The callback number must come from your own records, never from the message, the call or the meeting invite.
- A good protocol defines three things: the triggers, the verification action and the escalation path.
- The control only works if employees apply it under pressure, which is what simulation tests.
What does out-of-band verification mean?
Out-of-band verification means confirming who is behind a request by using a communication channel independent of the one the request came through. "Band" refers to the channel. If a request arrives by email, video call or phone, the confirmation happens somewhere else: a callback to a number on file, a message on an internal system, or a face-to-face check.
Two conditions make it work. The second channel must be pre-established, meaning its contact details were recorded before the request arrived. And the person making the request must not be able to steer you towards it.
How does out-of-band verification work?
The employee pauses the request, leaves the original channel and contacts the requester through a trusted route. A typical sequence:
In a May 2025 public service announcement, the FBI gives the same instruction for AI voice impersonation: research the person contacting you, "then independently identify a phone number for the person and call to verify their authenticity".
How is out-of-band verification different from out-of-band authentication?
Out-of-band authentication (OOBA) is a technical login factor: a code or push notification sent to a separate device to prove that the person signing in is the account owner. Out-of-band verification is a business process run by people to confirm that a request is genuine.
| Criterion | Out-of-band authentication | Out-of-band verification |
|---|---|---|
| What it confirms | The identity of a user at login | The legitimacy of a request |
| Who runs it | A system (MFA, push, SMS code) | An employee, following a procedure |
| Typical moment | Sign-in, transaction signing | Payment, bank detail change, access reset |
| Main threat covered | Stolen passwords | Impersonation by email, phone or video call |
Both are useful. Neither replaces the other: an attacker who convinces a finance employee on a call never needs to defeat MFA.
Why does out-of-band verification matter for enterprises?
Impersonation attacks win by controlling the channel. The attacker writes the email, places the call or hosts the meeting, so every check performed inside that channel is a check the attacker can answer. Moving to an independent channel breaks that control.
The financial stakes are well documented. Business email compromise caused almost $3.05 billion in reported losses in the United States in 2025, the second-largest category after investment fraud, according to the FBI IC3 annual report covered by CyberScoop in April 2026. Generative AI adds voice and video to the same playbook, so the BEC controls you already have must now cover calls and meetings too.
Detection tools help, but they score media rather than decisions. As covered in our review of what deepfake detection tools catch and miss, the final action is still taken by an employee.
What is a real example of missing out-of-band verification?
In early 2024, a finance employee at the engineering firm Arup in Hong Kong transferred HK$200 million (about US$25.6 million) in 15 transactions, as CNN reported in May 2024. He had first doubted an email from the "UK office". He then joined a video call where the CFO and other colleagues looked and sounded real. All of them were deepfakes.
The lesson is precise. The employee did seek confirmation, but inside a meeting the attacker had organised. A single callback to the CFO on a number from the company directory would have exposed the scheme, however realistic the faces were.
How do you set up an out-of-band verification protocol?
A protocol only works if it is short enough to apply under pressure. Define it in three parts.
Attackers anticipate the callback. In callback vishing, the phishing email itself supplies the number to call, which is why the number must always come from your own records.
- The list of triggering requests is written down and known by finance, HR, IT support and executive assistants.
- Contact details for verification come only from the directory, the ERP or the HR system.
- No callback number, link or meeting invite from the request itself is ever used.
- Payments above a set threshold require a second approver.
- Executives have agreed, in writing, that they will be called back and will not override the procedure.
- Employees know how to report a failed verification, and are thanked for doing it.
- The procedure is tested with realistic simulations, not only described in a policy.
How can you test whether employees apply it?
A policy shows what employees should do. A simulation shows what they actually do when a senior voice or face asks them to hurry. Running vishing and video call scenarios that end on a sensitive request measures how often employees stop and call back.
Arsen runs deepfake video call simulations that impersonate a customer's own executives on a pixel-accurate replica of the call interface. Arsen's team builds custom scenarios and environments for each company, department or internal profile: precise, workflow-specific scenarios for finance teams who approve payments, or broader scenarios that measure risk at scale across larger teams. Each simulation is paired with adaptive, interactive awareness lessons, and results are reported alongside phishing and vishing campaigns. The step-by-step walkthrough of a deepfake executive fraud simulation shows how a session runs.
Available platforms for deepfake video call simulations:
Microsoft Teams
Google Meet
ZoomRelated terms
- Impersonation: assuming a trusted identity to bypass verification.
- Business email compromise: email-based fraud that out-of-band verification is designed to stop.
- Vishing: voice phishing, including AI voice cloning.
- Out-of-band authentication: a login factor delivered on a separate device, covered above.