Skip to main content

FaceStealer: US Disrupts North Korea's Deepfake Meeting Crypto Scheme

FaceStealer is a North Korean social engineering campaign, attributed to the BlueNoroff group, that lured crypto, Web3, and open-source targets into fake video meetings using deepfakes and recorded footage of people they knew, then tricked them into running malware. The US government says it seized the campaign's cloud infrastructure; the scheme is linked to more than $400 million in stolen cryptocurrency.

On this page

What the US took down

The US government has disrupted a North Korean social engineering operation known as FaceStealer and seized the cloud infrastructure it ran on, after roughly 18 months of activity. BleepingComputer, reporting on 14 November, says the campaign is responsible for stealing more than $400 million in cryptocurrency, and that the FBI issued an alert about it earlier in the week.

In a separate announcement, the Justice Department said it and private companies had frozen more than $2.7 billion from the earlier Binance heist that the FBI tied to North Korean hackers, and had obtained court authorisation to seize or restrain more than $1.2 billion of it. That figure is not part of the FaceStealer total.

Security firm Arctic Wolf attributes FaceStealer to BlueNoroff, a financially motivated part of North Korea's Lazarus Group that has focused on crypto theft since at least 2014.

How the fake meetings worked

The scheme starts with trust. The attackers hijack accounts on Telegram, LinkedIn, Slack, and Discord and use them to invite a victim's contacts to a video call. On the call, the target sees what looks like people they know. Some are deepfakes, some are static profile photos, and some are recorded webcam footage of earlier victims.

Then the call breaks. A fake audio or technical error appears, and the target is told to run a quick fix, an update, an installer, or a command, which installs malware. Arctic Wolf found the operators also secretly recorded targets' webcams while they struggled with the broken call, and reused that footage to fool the next round of victims. US authorities say more than 900 webcam videos were captured this way.

The FBI's bulletin describes a maintainer of a popular JavaScript library, with tens of millions of weekly downloads, who was lured into a fake Teams meeting before the attack was caught, and other attempts against npm and Rust developers. A compromised maintainer can turn one stolen laptop into a supply chain attack.

Off-the-shelf AI made the deepfakes cheaper

Arctic Wolf's 12 November report shows how fast the production side matured. In April, the firm documented the group's AI-generated fake Zoom meetings; at that stage, its new report says, a single editor was using Adobe Premiere to make basic deepfakes. By November, BlueNoroff was combining legitimate tools including Google generative AI, Grok Imagine, Wondershare AIGC, and GPT-4o to make avatars and synthetic participants, and was screen-recording YouTube interviews and conference talks to capture people's faces and voices.

The target list grew with it, from more than 400 people across 20 countries in April to 677 across 58 countries, 84% of them founders and senior executives. Arctic Wolf flagged 349 videos in the group's media library as likely modified, and stresses that it identified deepfakes from provenance evidence, not by eye. That is the uncomfortable lesson: a convincing face on a call is no longer proof of who you are talking to.

How to protect yourself

Arctic Wolf's core advice is simple: treat any request to run a command, script, or installer during a call as a red flag, however familiar the faces look. Confirm unexpected meeting invites and identity checks through a separate channel you already trust, be wary of helpful PDFs and video links from outside contacts, and use strong authentication for anyone with access to funds or code.

The webcam footage is part of the threat too. If a call fails and asks you to fix it, leave the call. Covering or disabling your camera until you have verified the meeting stops your face from becoming the next lure.

CuriousLM runs supported AI models locally on your device. Try CuriousLM.