Skip to main content
Diopter
Log inBook a demo
Capability

Real-time social engineering defense for Zoom, Teams, and Meet

Diopter helps your team spot a deepfake or a risky request while the Zoom, Teams, or Meet call is still live. We test your people with realistic social engineering simulations, then assist them on real calls with a real-time alert before a payment, a password reset, or an access change goes through.

Your first agentic attack is on usBook a walkthrough

A free simulated attack on a team you choose, with an NDA first if you want one.

What can Diopter help your team do on live calls?

Diopter helps your team confirm that the people on a live call are who they say they are, and that their face and voice are real, before money moves or access is granted.

Where it helps

  1. Approve payments on video safely. A cloned executive or vendor joins the call to "confirm" a transfer or new bank details.

  2. Reset access without falling for a fake. An "employee" sends a message on Teams, locked out and needing an MFA reset in the next 5 minutes. Or a fake help desk agent calls your staff.

  3. Protect client and investor calls. Your client gets a call from "you," or you get one from "them," looking for money, data, or both.

  4. Prove your controls to auditors and insurers. If they ask how you verify payment and access requests made on calls, simulation results and the evidence trail from live calls can help you answer.

How does Diopter help stop deepfake calls on Zoom, Teams, and Meet?

Diopter checks 4 things on every call, from the first second until everyone has left.

Identity

What Diopter checks

Whether the person and the request match your records, and whether the payee is someone you have paid before. Diopter reads the invite and the email thread as context, so it already knows about a lookalike domain or a first-time contact when the call connects.

AlertMismatch

Voice and video

What Diopter checks

Cloned or text-to-speech voices, even one that only turns up halfway through, plus face-swap, reenactment, and lip-sync artifacts on video. Everyone on the call gets their own score.

AlertSynthetic

Conversation

What Diopter checks

Claimed authority (“I’m the CFO”), a deadline (“it has to be today”), or a move to a new channel (“keep it off email”).

AlertPressure rising

Policy

What Diopter checks

The request against the steps your approval process requires.

AlertOut of policy

The person on the call sees the alert, the reason, and a suggested next step, such as holding the request. Security gets the evidence in the console, so a follow-up message on another channel arrives already flagged.

Which deepfake attacks have already hit live calls?

Arup, WPP, and Ferrari were all targeted on live calls by attackers using a cloned executive's face or voice.

  • In early 2024, a finance employee at the engineering firm Arup joined a video call with what looked like the CFO and several colleagues, then approved 15 transfers worth about US$25 million. It turned out that none of the colleagues on the call were real.1

  • Scammers targeting WPP set up a Microsoft Teams meeting through a WhatsApp account that used CEO Mark Read's photo. They used a voice clone built from YouTube footage to ask an agency leader to start a new business. The attempt failed.2

  • In July 2024, a Ferrari executive got WhatsApp messages and then a call from a cloned voice of CEO Benedetto Vigna about a confidential deal. When the executive asked about a book Vigna had recommended days earlier, the caller hung up.3

These attacks follow the same pattern:

  1. 01

    A familiar face or voice joins. It looks and sounds right.

  2. 02

    It has to happen today, because nobody verifies when there’s deadline.

  3. 03

    A request for discretion keeps the conversation away from anyone who would push back.

  4. 04

    The ask grows, and the money moves once the call ends.

Why do deepfake attacks work on live calls?

A few minutes of someone's public footage from an earnings call or a webinar is now enough for off-the-shelf software to borrow their face and voice on a live call. We hear "I'd spot it" on almost every first call, usually from smart people, but the Ferrari executive only caught the fake by asking a question the clone couldn't answer. Small firms get hit too, since "we all know each other here" is the gap a fake identity slips through. The FBI recorded US$2.77 billion in business email compromise losses in 2024 and has warned that criminals now bring AI video and audio onto live calls to pass as executives and vendors.4

How do Diopter's social engineering simulations test your team?

Diopter runs realistic simulated deepfake calls on the people attackers go after first: finance, the IT help desk, executive assistants, and new joiners. The caller can use a cloned executive's voice or join as a deepfake on video. You see who held and where the attack worked, so you can fix your policies and train your people. Do the dry runs with us so our agents hit you before a real attacker does.

  • We’ll run simulated Zoom, Teams, and Meet calls built from real attacks, such as payment approvals, MFA resets, and executive impersonation.
  • You’ll receive results by team and by step, so you know which teams to train first.
  • The same scenarios then become the checks Diopter will run on live calls.

Which platforms does Diopter's real-time deepfake detection run on?

Diopter's real-time deepfake detection runs on Zoom, Microsoft Teams, Google Meet, VoIP, and phone calls, with Webex coming soon.

There's nothing to upload and no change to how people take calls.

Zoom

How Diopter runs

Joins as a participant or runs bot-free from the user's device

Microsoft Teams

How Diopter runs

The same 2 options, rolled out through your tenant and MDM

Google Meet

How Diopter runs

Runs live on the call with the same checks

VoIP and phone

How Diopter runs

Scores voice and conversation where there is no video

Webex

How Diopter runs

Coming soon

How does Diopter compare with other ways to stop deepfake calls?

Diopter works during the live call and checks what the caller is asking for. File detectors, voice biometrics, and training each miss at least one of those.

A file detector tells you the video was fake after the money is gone. A good clone can beat voice biometrics. Training only works if someone spots the signs mid-call with their boss's face on screen. And none of them checks the request itself.

  • Works live on the call

    File detectors
    No
    Voice biometrics
    Yes
    Awareness training
    No
    Diopter
    Yes

  • Detects synthetic video

    File detectors
    Yes
    Voice biometrics
    No
    Awareness training
    No
    Diopter
    Yes

  • Detects cloned voice

    File detectors
    Yes
    Voice biometrics
    Partly
    Awareness training
    No
    Diopter
    Yes

  • Reads the conversation for pressure

    File detectors
    No
    Voice biometrics
    No
    Awareness training
    Relies on the person
    Diopter
    Yes

  • Checks the request against your policy

    File detectors
    No
    Voice biometrics
    No
    Awareness training
    No
    Diopter
    Yes

How does Diopter roll out for a security team?

Diopter rolls out in two steps. Pilot with one team in days, then go wider through MDM once you're satisfied.

  • OpenTelemetry native: logs flow into your SIEM, and EDR and MDR connect via API

  • Rolls out through MDM (Intune, Jamf)

  • Bot or bot-free capture

  • Configurable data retention, including zero retention

  • On-prem and hybrid deployments, so call media stays inside your network

  • Nothing to install on the caller's side

  • SOC 2 Type II completed

Two ways to start

Your first agentic attack is on us. Fill out a form and we'll chat about how to run it for one of your teams. Want to talk about defending live calls instead? Book a walkthrough, and we'll replay a real deepfake call and show you what Diopter caught and when.

Your first agentic attack is on usBook a walkthrough

Common questions for security and finance teams

Can Diopter test our team with a simulated deepfake call first?

Yes. Diopter runs simulated calls on Zoom, Teams, and Meet against finance, the help desk, and other teams, with a cloned voice or deepfake video if you want it, and shows where each team held. Most teams start there, then turn the same scenarios on for live calls.

Does real-time deepfake detection work on Google Meet, Webex, and plain phone calls?

Google Meet and plain phone calls, yes, and Webex is coming soon. Diopter works on Google Meet as well as Zoom and Teams, and the checks are the same everywhere. On VoIP and conference lines, through Dialpad, RingCentral, or any carrier via API, and anywhere there's no video, we listen for cloning in the voice and for pressure in what's said. You get the same real-time alerts and confidence level as on video.

Does Diopter join the call as a participant?

Diopter can join as a participant, but it doesn't have to. In bot mode it joins under a name everyone on the call can see. In bot-free mode it runs on your team member's device and the call looks the same as usual.

How does Diopter avoid false alarms on busy calls?

Diopter won't flag a normal call just because someone's in a hurry. Your finance lead chasing a real month-end close is fine, because one glitch or one pushy sentence doesn't raise an alert. We look for the pattern of claimed authority, real pressure, and a request that keeps getting bigger.

What should you look for in a real-time deepfake detection tool?

Start with your own process. A callback on a known number and a second approver for payments, resets, and access requests stop many deepfake calls without any tools. When you compare real-time deepfake detection tools, run each one on your own meetings during a pilot and measure how often it raises a false alarm, how fast the alert reaches the person on the call, and what it does with recordings and data. Check that it covers the meeting platforms and phone lines your team uses and fits your existing security tools. The final decision should stay with a person. Your own pilot will tell you more than any vendor's benchmark, including ours.

Research behind real-time deepfake detection

  1. CNN, May 16, 2024, on the Arup deepfake scam.
  2. Storyboard18, May 13, 2024, citing The Guardian, on the WPP attempt.
  3. Fortune, Jul 27, 2024, citing Bloomberg, on the Ferrari attempt.
  4. FBI IC3 2024 Internet Crime Report, and FBI PSA I-120324-PSA, Dec 3, 2024.