The next crypto scam invitation could arrive from someone you already know, with their real account, familiar messages and a video of their face. FaceStealer turns those signs of trust into tools for theft. The FBI says the North Korean campaign stole more than $400 million over 18 months and used a library of stolen webcam recordings.
The FBI’s FaceStealer warning describes a repeating cycle: compromise one person, take over their communications, then use recorded footage to approach their contacts. Authorities announced an infrastructure disruption in November 2026 and said they seized more than 900 victim videos.
The important finding is not simply that a convincing fake call can fool someone. One compromise can supply the account, relationships and visual material for the next approach. Comparing the FBI’s account with researchers’ technical findings shows why checking a sender’s profile or recognizing a face may leave the central risk unresolved.
How FaceStealer turns trusted contacts into new targets
According to the FBI’s November 12 technical advisory, attackers invite contacts to imitation meetings, play familiar people’s footage and introduce a supposed technical problem. Instructions to repair it cause the target to install malware. Stolen communications and recordings then support further approaches.
This attacks a reasonable habit: people treat an established relationship as evidence that a request is genuine. A hijacked account changes that calculation. The identity of the account’s owner and the identity of its current operator are separate questions.
The practical distinction is between recognizing someone and verifying a request. A familiar face does not establish that a software installation, wallet prompt or computer command came from that person.
Inside the fake meeting sits an operator console
JUMPSEC’s source-code investigation provides a closer view of the machinery. Researchers say exposed source maps, files that help reconstruct website code, revealed Zoom- and Teams-themed kits. Their analysis describes wallet detection, operator controls and staged technical-error messages.
In the analyzed kit, wallet information could reach an operator before malware delivery. The researchers also found a clipboard trap: apparently harmless displayed text could be replaced with a malicious command when copied. These are findings about the examined variants, not proof that every FaceStealer attempt uses identical code.
The sequence explains why a polished interface is a weak safety test. Reactions, settings panels and familiar meeting layouts can serve the deception. For the target, the decisive moment is the request to take an unusual action on their computer.
JUMPSEC documented both Windows and macOS chains. Its report also states that some later-stage payloads were not recovered. That gap matters: a detailed reconstruction can establish parts of an attack without exposing every capability deployed against every victim.
Why the campaign now reaches beyond video calls
In research published November 12, Arctic Wolf describes expansion into shared financial documents, fake PDF-viewer updates, media players, identity checks, wallet access and security workflows. Different packaging leads toward the same demand: perform a supposed fix.
The firm identified material concerning 677 people across 58 countries in the attackers’ media library. Founders and senior executives represented 84% of that set. Arctic Wolf explicitly says these are people targeted with prepared material, not a count of confirmed compromises.
Its researchers also distinguished AI-generated material from footage altered with conventional editing tools. They relied on evidence of how media was produced, rather than judging appearance alone.
For readers, that undermines a comforting assumption: that the defense is learning to spot an awkward face or unnatural movement. The stronger question concerns the action being requested, even when the person on screen appears convincing.
What the headline numbers actually tell us here
| Reported figure | What it measures | What it does not establish |
|---|---|---|
| More than $400 million | Campaign thefts reported by the FBI over 18 months | Money recovered in the infrastructure disruption |
| More than 900 videos | Victim webcam recordings seized by the FBI | 900 separate thefts or 900 unique people |
| 677 people | People identified in Arctic Wolf’s examined media library | 677 confirmed compromises |
These figures should not be added together or used to calculate an average loss per victim. They describe different units and evidence sets. Neither the records reviewed nor the public totals establish a complete list of individual losses.
This is also separate from the Binance asset-seizure action announced alongside FaceStealer. BlockBriefing’s earlier report on that announcement covers the distinction. A shared announcement does not make the incidents one theft.
A fake update can defeat familiar security assumptions
Microsoft’s Sapphire Sleet research, originally published in April and updated in June, documents macOS intrusions driven by deceptive updates and user execution. Attackers persuaded people to run files that led to credential harvesting and other malicious activity.
Microsoft says Apple added protections against associated infrastructure and malware. It still recommends that users never execute scripts or commands supplied through messages or calls without approval from their IT or security teams.
This older research supplies technical context for the November warnings. It is not a new November incident, and its findings should not be treated as a complete description of every variant.
For a crypto newsroom, founder or investor, the relevant boundary is straightforward: a meeting participant should not become an unofficial software-support authority merely because they are recognizable.
What readers can do before the next invitation
JUMPSEC recommends checking requests through a separate communication channel before interacting, even when the invitation comes from a trusted contact. Microsoft’s guidance supports stopping when an unexpected meeting asks for command execution or software installation.
- Verify unusual requests through an independently established contact route.
- Stop a meeting that asks you to run a script or install an unexpected update; involve your security team.
- Keep devices updated through official mechanisms rather than links supplied during a call.
For related controls, read BlockBriefing’s guide to protecting exchange API keys. Credential hygiene complements request verification; it does not make a suspicious meeting safe.
What this evidence establishes and what remains unknown
The combined record supports a specific conclusion: attackers are exploiting trusted relationships and routine work, with manipulated media helping the approach. It does not prove that every convincing video call is malicious or that the November disruption ended the campaign.
Reporting method: This AI-assisted report compares public FBI materials with original research from Arctic Wolf, JUMPSEC and Microsoft. Attribution, loss estimates and technical findings remain credited to those sources. BlockBriefing did not access criminal infrastructure, inspect private victim records, test malware or conduct interviews.
