The short version: The FBI attributes FaceStealer to North Korean hackers and says the scheme has been linked to more than $400 million in cryptocurrency theft over 18 months. Investigators seized more than 900 webcam videos. The campaign primarily targeted people in cryptocurrency and Web3, with attempts against open-source software maintainers.
What to do: Verify unexpected meetings through a separate trusted channel, open your meeting app independently, and do not run a command or install a tool because someone in a call says your audio is broken.
FaceStealer: Fake Video Calls, Real Faces and $400 Million Stolen
The FBI says North Korean hackers used fake video meetings and recordings of real people in a scheme that stole more than $400 million in cryptocurrency. The most unsettling detail is how the attackers turned each compromised relationship into a way to reach someone else.

Jump to a section
The invitation can come from an account you recognize. The face on the screen can belong to someone you know. Then the sound fails, a message suggests a quick fix, and what looks like a routine meeting becomes a security incident.
That is the chain described in a November 12 FBI warning about FaceStealer. The bureau says the operators used hijacked accounts, counterfeit meeting pages and stolen webcam recordings to persuade targets to install malicious software. The fake technical problem is the moment when familiarity becomes leverage.
Those figures are the FBI's account of this campaign, not a measure of all fake-meeting scams. The Justice Department announced the disruption of infrastructure used by the operation. That does not establish that every related attack has stopped. We checked the sources on November 16.
How a familiar face becomes the bait
The attack begins before the invitation arrives. According to the FBI, the operators first compromise a person's device and communication accounts, then collect contacts, chat history and webcam footage. From a hijacked account, they approach people who have reason to know the original victim.
The next target may receive a scheduling invitation for an ordinary-looking meeting. Shortly before it starts, the attackers send a link to a lookalike meeting site under their control. A genuine account and a plausible calendar invitation can make the destination feel more trustworthy than it is. Neither proves that the meeting link is safe.
In the fake meeting, the attackers play recorded video of a known contact, sometimes alongside footage of other supposed participants. A message claims the target cannot be heard. The proposed solution, an update, script or command to paste, installs malware if followed. The operator can then use the new victim's access and recordings to approach more people. The full sequence and its limits are described in the FBI's FaceStealer overview.
- A familiar account is hijacked
- A fake meeting invite arrives
- Recorded faces build trust
- A fake fix installs malware
- More contacts are targeted
Hacked.com explanation based on the FBI's reported pattern. Not every invitation results in an infection.
There is a difference between receiving an invitation, opening a site and installing malicious software. They do not all mean the same thing. If you only received a message, verify it through a trusted channel. If you ran a supposed fix, treat the device as potentially compromised and follow the response steps below.
The tactic did not appear overnight. In a separate 2025 investigation, security company Huntress described a cryptocurrency foundation employee being drawn into a fake meeting with apparent colleagues and told to download a supposed Zoom extension. Huntress attributed that intrusion to BlueNoroff with high confidence. It is historical context for the technique, not evidence that the same victim appears in the FBI's seized video collection.
The deepest trick is not necessarily a deepfake
AI-generated impersonation is part of the story, but it is not the whole story. The FBI says the operators have begun using synthetic videos of prominent people in crypto and Web3. Its advisory also says the scheme still relies overwhelmingly on genuine recordings taken from previous victims.
That distinction matters. A real video can be used deceptively: the person in it is genuine, but the apparent conversation is not. Looking for an unnatural face or a strange voice may miss the problem entirely. A better question is whether the request itself makes sense and whether the meeting was confirmed through a channel you control.
Arctic Wolf's November research adds another piece. The company says the group it tracks as BlueNoroff expanded its manipulated-media workflow and used generative AI in producing synthetic participant videos. Its analysts say they used evidence about a video's origin and production, rather than appearance alone, to identify deepfakes.
Those observations and the FBI advisory cover related activity, but they are not the same dataset. Arctic Wolf identified material concerning 677 individuals across 58 countries in the actor's media library; it explicitly says that count reflects people for whom material was prepared, not confirmed compromises. The FBI's more than 900 seized videos are a different measure. They should not be added together as a victim count.
When your face becomes someone else's disguise
There can be more than one victim in an impersonation. The person receiving a fraudulent request may lose money or account access. The person whose face was used can also face suspicion, distress and repeated attempts to clear their name.
In a separate Associated Press investigation, Hunter Bigham described scammers using his photographs for false romance profiles and people deceived by those accounts contacting him. That reporting is about romance scams, not FaceStealer; there is no established connection between the operators or the victims. The shared problem is that a genuine image can become a tool for someone else's deception.
If you learn that someone is using your image, document the fake profile or message, report it through the platform's impersonation process, and warn relevant contacts through a channel you control. Avoid confronting someone who may have been deceived as though they knowingly assisted the scam. Our romance scam and threats guide explains how to preserve evidence and respond when intimidation follows.
The fake meeting is becoming a fake document
Arctic Wolf found that the same group had expanded beyond meeting invitations into supposed shared financial documents, PDF viewer updates, identity checks and other familiar workflows. Its account describes a recurring pattern: establish professional trust, create a technical obstacle, then offer instructions that make the target run the attacker's fix.
That is the connection that makes the warning useful beyond one brand or application. A fake meeting is one wrapper. A document that supposedly will not open can be another. The decisive step is the request to install software, enter credentials or execute a command that the recipient did not independently verify.
These observations come from attacks against particular crypto and professional targets. They do not establish that the identical FaceStealer operation is targeting every household. But the underlying technique is familiar in fake technical support and remote-access scams: an invented problem is used to make an unsafe instruction feel like routine help.
One compromised developer, many possible downstream victims
The FBI describes a separate reason the campaign matters: attackers also sought access to open-source software maintainers, the people who update code used by other organizations. In one case, the bureau says an attacker persuaded a maintainer of a widely downloaded JavaScript library to follow fake corrective instructions during a counterfeit meeting.
The FBI says the attackers then used the maintainer's access to create malicious dependencies that would have been downloaded with the legitimate library. It credits quick detection and response with preventing a potentially much wider compromise. The advisory does not name the library or establish that its users were infected, so we cannot identify it or count downstream victims.
The contrast is important: the immediate victim is one person at a computer; the possible consequence is the distribution of altered software to many others. A business can do everything right in its own meeting and still depend on code maintained elsewhere. That is a reason for organizations to review update provenance and limit permissions, not a reason to assume every open-source package is unsafe.
The danger can return months later
The FBI says that in several reported cases, attackers waited months or more than a year after the initial compromise before stealing cryptocurrency or taking over communication accounts. A few quiet weeks after a suspicious call are therefore not proof that a device or wallet is safe.
Cryptocurrency wallets are especially sensitive because a copied recovery phrase or private key may allow someone else to control the assets without access to the original device. Changing an email password does not invalidate a copied wallet recovery phrase. If you believe a wallet's secrets were exposed, seek qualified help promptly and use a separate, trusted device to arrange a safe transfer. Never give anyone your recovery phrase or private key.
The Justice Department announced the FaceStealer infrastructure disruption alongside an action involving a separate Binance theft. Its November 12 announcement treats them as separate matters. The figures and facts of one case should not be folded into the other.
How to check a suspicious invitation
- Confirm the meeting independently. Call or message the person using a number or channel you already trust. A reply in the same possibly hijacked account is not independent confirmation.
- Open the meeting service yourself. Use the official app or type the provider's known address yourself, then enter the meeting details. A link in a calendar invitation or chat is not automatically safe.
- Keep the camera off while you verify. The FBI recommends waiting until you can see, hear and communicate with the inviter. A live exchange alone is still not proof of identity; confirm unusual requests separately.
- Do not follow an in-call repair command. Leave the meeting if asked to paste a command, run a script, install an unfamiliar tool or bypass a security warning. Check for updates inside the official app or through the provider's verified website.
- Protect your communication accounts. Use unique passwords where applicable, turn on available multifactor authentication and login alerts, and check recovery details and active sessions.
For more on manipulated video and practical verification, see what deepfakes are and why they are dangerous. A recognizable voice, familiar profile picture or polished message should prompt the same verification as any other unexpected request involving access or money.
What to do after a suspicious meeting
Your next step depends on what you did. An invitation, a browser permission and an installed program are different exposures; none alone proves that your money or every account was accessed.
- Received an invitation but did not open it: Do not use the link. Check with the contact through a known phone number or another trusted channel, and report or delete the message as appropriate.
- Opened the page or allowed your camera: Leave the page. Review and remove the site's camera and microphone permissions in your browser settings, where available. A camera permission does not by itself prove malware was installed, but footage already captured cannot be taken back. Warn relevant contacts if your image may be misused.
- Entered a password or approved a sign-in: From a different trusted device, use the provider's official site to change the password, review recovery methods and active sessions, revoke unfamiliar access and turn on multifactor authentication. Contact the provider if you cannot sign in.
- Downloaded a file but did not open it: Do not open it. On a personal device, remove it if safe; on a work device, follow your IT team's instructions, including any evidence-preservation request. If you are unsure whether it ran, treat the device as potentially compromised.
- Ran the fix or installed unfamiliar software: Disconnect the device from the internet and follow the containment steps below. Contact your IT or security team promptly if it is a work device.
If you installed the supposed fix
Disconnect the affected device from the internet and stop using it for sensitive accounts. If it is a work device, contact your organization's IT or security team immediately and follow its instructions. Do not rush to erase the device before discussing evidence and recovery with the people investigating.
From a different device you trust, secure your primary email and the communication account used for the invitation. Change passwords, review recovery methods, end unfamiliar sessions and enable multifactor authentication where available. If a password manager or connected applications may have been exposed, investigate those separately. Changing one password does not necessarily revoke every session or application credential.
Tell affected contacts through a separate trusted channel that messages or video bearing your identity may be misused. Contact your bank, payment provider or crypto exchange through verified channels if money or assets may be at risk. The FBI asks people reporting suspected FaceStealer activity to use IC3 and include the keyword FaceStealer. Readers outside the United States can also report to their local cybercrime or law enforcement service.
Our account recovery guide explains how to work out which recovery channels you still control. For organizations, our reporting on stolen AI keys and connected credentials examines a related but separate problem: access held by software can survive a person's password change.
Trust the process, not the picture
FaceStealer turns the evidence people rely on for trust against them: a known account, a familiar face, a scheduled appointment and a plausible technical problem. None has to look obviously fake for the chain to work.
The practical defense is to separate recognition from authorization. You can recognize the person on a screen and still verify the meeting. You can want to fix the sound and still refuse to run a command. Those small pauses interrupt the attack at the point where the attacker needs your cooperation.