← All articles

North Korea on both sides of the interview: the fake recruiter behind the $6.3 billion Binance theft, and the fake applicants at Kraken and in Japan

North Korean operatives pose as recruiters to get into engineers' laptops and as applicants to get onto payrolls. How it worked at Binance and Kraken, the Japanese cases, and how to stop it.

Incident period: Binance: June – 29 Oct 2026; Kraken: 2025; Japan: 2024–26Sources checked: 4 Nov 2026

30-second briefing

Key facts
Binance says an engineer downloaded a fake interview's 'assignment' onto his work computer in September. With an earlier tampered installer and an internal AI tool, the attacker reached wallet signing and took about $6.34 billion on 29 October; the FBI blames North Korea. Kraken and a Japanese exchange each caught a North Korean posing as an applicant.
Who should act
IT Operations · EDR · Legal / Privacy · Third-party Risk
First action
Tell staff today: no job tests, take-home tasks or 'meeting fixes' on work devices. Ask HR to keep cameras on, ask live local questions and check ID against the person, and send laptops and salaries only to verified addresses and accounts in the employee's own name.
Uncertainty to keep in mind
Binance's findings are preliminary, and it relies on the FBI for attribution. Our research network could not open the companies' and authorities' pages, so we read them through search extracts. Kraken has not named its applicant, and Japan's exchange is unnamed.
Header graphic: Both sides of the interview. Four panels: $6.34 billion taken from Binance in an attack that ran through a fake job interview; 30,000+ PCs infected by fake interviews; 187 laptops and KVM devices seized in the US; Japan's first laptop farm found.
THE POSTURE QUESTION

North Korea uses hiring as a way in from both sides. As a fake recruiter, it gives your engineers a coding test that runs its malware on their laptop; at Binance that chain ended in a $6.34 billion theft. As a fake applicant, it borrows an identity, gets through a coached video interview and asks for the laptop to go to a helper's home. Two habits close most of both doors: never run a stranger's code on a work machine, and never let a document stand in for a live, unscripted conversation.

North Korea has learnt to use hiring as a way in, from both sides of the interview. Posing as a recruiter, it offers your engineers a job and asks them to run a coding test that installs its malware. That is how Binance says an attacker got onto an engineer's work computer in September 2026, the second of four stages that ended on 29 October with about $6.34 billion of digital assets taken, a theft the FBI attributes to North Korea. Posing as an applicant, it borrows someone's identity, gets through a coached video interview and, once hired, asks for the company laptop to be sent to a helper's home. Kraken caught one such applicant in 2025; a Japanese crypto exchange caught another. This article is built mostly from pictures, for the people who hire and the people who get hired: what happened, how each scheme works, the red flags, and which team can stop it. For North Korea's wider aims, see our analysis of the four state actors.

What happened, and what is confirmed

Status as of 4 November 2026. "Which door" refers to Figure 1 below: Door 1 is the fake recruiter, Door 2 the fake applicant.

Case Status What happened
Binance, June – October 2026 · Door 1 Confirmed by Binance; the FBI attributes it to North Korea An engineer downloaded an interview "assignment" onto his work computer. With a tampered installer from June and an internal AI tool, the attacker reached wallet signing and took about $6.34 billion on 29 October.
DMM Bitcoin via Ginco, 2024 · Door 1 Attributed by the FBI and the NPA A "recruiter" on LinkedIn sent a Ginco employee a Python "pre-employment test". The attackers later altered a DMM transaction: 4,502.9 BTC, about $308 million (about ¥48.2 billion).
WaterPlum ("Contagious Interview"), Dec 2025 – Jul 2026 · Door 1 Attributed by Japan, the US, Australia and Germany Fake employers asked developers to run coding tests. At least 30,000 PCs in more than 100 countries were infected and over 7,000 crypto wallets emptied or their credentials taken.
Kraken, 2025 · Door 2 Kraken's own account; not hired Kraken spotted a North Korean applicant, kept them in the process to study them, and watched them fail live checks in the final interview.
A Japanese crypto exchange, May 2025 · Door 2 Reported by the NPA; not hired An applicant for an engineering job posed as someone else. The company noticed oddities in the interview.
Laptop farms in Japan and the US · Door 2 Reported by the NPA and the U.S. Justice Department Japan's first laptop farm for North Korean IT workers was found, with several hundred million yen sent abroad. In Texas, 187 laptops and KVM devices were seized.

NPA is Japan's National Police Agency. "Attributed" means the authorities named North Korea; it is not a court finding.

Two doors into the same company

Diagram of two parallel flows that meet in one box. Door 1, the fake recruiter, comes to your engineers in five steps: a message on a real job platform or LinkedIn; video interviews with "HR", then a "CEO"; a coding test or repo to download and run; malware on the laptop, often a work laptop; and access to credentials, wallets and signing systems. Its documented cases are Binance in 2026, DMM Bitcoin via Ginco in 2024, and the WaterPlum campaign with more than 30,000 infected PCs. Door 2, the fake applicant, comes to your HR team in five steps: a borrowed or stolen identity and an AI-polished CV; coached video interviews, sometimes with a face swap; once hired, the laptop goes to a helper's home; a laptop farm controlled from abroad; and salary, data and insider access. Its documented cases are Kraken in 2025 and a Japanese exchange in 2025, both caught, and 187 devices seized in the United States. Arrows from both doors lead to a box labelled "Same paymaster": the NPA and the FBI assess that the fake-recruiter group WaterPlum and some North Korean IT workers both operate under the 313 General Bureau of the Workers' Party's Munitions Industry Department, and the NPA says the workers' earnings fund nuclear and missile programmes.

Figure 1. Original graphic built from Binance's, Kraken's and the authorities' own accounts.

The two doors are run by the same state. Japan's NPA and the FBI assess that WaterPlum, the group behind the fake interviews, and some North Korean IT workers both operate under the 313 General Bureau of the Munitions Industry Department of the Workers' Party's Central Committee. Some WaterPlum operators also work as IT workers from the same internet addresses. One door steals; the other earns salaries and opens a door from the inside. The NPA says the IT workers' earnings fund North Korea's nuclear and missile programmes.

For a general reader: if a stranger offers you a job and then asks you to run their software, assume it is an attack. If a candidate will not keep their camera on, cannot talk about the city they say they live in, or wants their laptop sent somewhere else, assume the same.

Door 1: the recruiter who isn't

Binance: four stages, three engineers

Swimlane chart of Binance's account, with rows for Engineer A in wallet DevOps, Engineer B a software engineer, Engineer C in wallet software, and Binance's security team, and columns for June 2026, September, 20 October, 27 October and 29 October. In June, Engineer A searches the web for a popular AI coding tool and installs a tampered Homebrew package; Binance's security team detects it and replaces the device, but Engineer A's role and wallet colleagues are now known to the attacker. In September, a "Project Director" on a real job platform contacts Engineer B; after interviews with "HR" and then a "CEO", Engineer B downloads the "assignment" repo onto his work computer. On 20 October the attacker works from Engineer B's computer and uses an internal AI tool to gain control over Engineers A and C. On 27 October Binance says it blocked several North Korea-linked fake-interview attacks in the past week, with no losses. On 29 October stolen deployment credentials on Engineer C's computer launch an unauthorised container that reaches wallet signing. Below, four figures: $6.34 billion impact estimated by Binance, with the FBI saying more than $6 billion; 98% of affected token assets replenished by 31 October; 57 suspected North Korea-linked cases Binance says it stopped in 2026; and the FBI's attribution of the theft to TraderTraitor, North Korea, on 31 October.

Figure 2. Original graphic built from Binance's incident updates and the FBI's alert. Columns are not to scale.

Binance's own account describes three engineers and four stages.

  1. June: a tampered installer. Engineer A, who works on wallet operations, searched the web for a popular AI coding tool and installed a tampered Homebrew package along the way. Binance's security team detected it and replaced the device. But, in Binance's words, Engineer A's role and his colleagues on the wallet infrastructure team were now exposed to the attacker.
  2. September: the fake recruiter. Engineer B was contacted on a legitimate job-hunting platform by someone posing as the Project Director of a fictitious company; his profile mentioned his Binance experience. A first-round interview with an "HR representative" was followed by a second with the company's "CEO", who shared personal social-media content to build trust. In the second round he was told to download an "assignment" containing a malicious code repository, and he did so on his work computer.
  3. 20 October: an internal AI tool. Working from Engineer B's computer, the attacker used an internal AI tool to gain control over Engineers A and C. Binance has since rebuilt that tool and its access controls.
  4. 29 October: the theft. Binance's first update said a compromised DevOps engineer's laptop let the attacker reach a service container and submit unauthorised withdrawals, and that it had found no evidence that any wallet private key had leaked. The later update says the attacker reached an internal wallet-signing service and used Ethereum's EIP-7702 authorisation feature, outside Binance's normal withdrawal infrastructure.

The damage and the response. Binance estimates the impact at about $6.34 billion and says that by 31 October it had replenished 98% of the affected token assets, verifiable on its proof-of-reserves page. It also reports about $2.85 billion frozen or intercepted. On 31 October the FBI attributed the theft of "more than $6 billion" to North Korea's TraderTraitor, saw the funds being converted mainly into Ether, and asked exchanges and other services to block the laundering addresses. Binance relies on that attribution rather than making its own, and says its findings will be reviewed independently.

The detail that matters most for everyone else. Binance was not careless about this threat. It says it stopped 57 suspected North Korea-linked cases in 2026, and two days before the theft it said it had blocked several fake-interview attacks the previous week with no losses, as Wu Blockchain relayed. One engineer running one "assignment" on a work computer was enough, because that computer could reach tools that reached the wallets.

How a fake interview works

Six numbered cards in two rows, each with a description and, in green, the moment a target could have stopped it. 1, pick the target: profiles that name the employer and role, as Binance's engineer did on the job platform; stop when an approach singles out your current job. 2, make contact: a "Project Director" on a legitimate job site at Binance, a recruiter on LinkedIn at Ginco, and posing as AI, crypto or NFT firms in WaterPlum; stop at an unsolicited offer from a company you cannot verify. 3, build trust on video: a first round with "HR" and a second with the "CEO" of a fictitious company, who shares personal social media to seem real; stop at quick escalation to someone senior and charming. 4, hand over the test: download an "assignment" repo at Binance, a Python "pre-employment test" on GitHub at Ginco, or a coding test or a "fix" for a video-call error in WaterPlum; stop at any request to run their code or install their tool. 5, run the payload: malicious npm packages carrying BeaverTail, InvisibleFerret, OtterCookie or OtterCandy, or a task hidden in a VS Code project; stop before doing it on a work computer. 6, pivot to what matters: stolen session cookies into a messaging system at Ginco and then DMM, and an internal AI tool and deployment credentials to wallet signing at Binance; stop by making sure signing systems are not reachable from an ordinary laptop.

Figure 3. Original graphic. The green boxes are the moments a target, or the company around them, could have stopped the chain.

Japan saw this first at DMM Bitcoin. In late March 2024, according to the FBI and the NPA, someone posing as a recruiter on LinkedIn contacted an employee of Ginco, a Japanese company that makes wallet software for businesses. The "pre-employment test" was a malicious Python script on GitHub, and the employee ran it. From mid-May, the attackers used the employee's stolen session cookies to get into Ginco's unencrypted messaging system. In late May they altered a legitimate transaction request from a DMM employee, and 4,502.9 BTC, worth about $308 million at the time, went to wallets they controlled.

Since then it has become a mass campaign. On 18 September 2026, Japan's NPA and National Cybersecurity Office, the FBI, the U.S. Defense Cyber Crime Center, Australia's Cyber Security Centre and Germany's BND and BfV publicly attributed WaterPlum, also called Contagious Interview (joint advisory, PDF; NPA). Posing as employers, often AI, crypto or NFT firms, the group asks developers to complete a coding test or to fix an error in a video-call tool. Running it installs malicious npm packages carrying BeaverTail, InvisibleFerret, OtterCookie or OtterCandy, or runs a task hidden in a VS Code project. From December 2025 to July 2026 it infected at least 30,000 PCs in more than 100 countries, including Japan, and moved at least ¥1.7 billion ($10.71 million) out of more than 7,000 wallets.

Door 2: the applicant who isn't

Kraken: interviewing a North Korean on purpose

A five-stage hiring funnel from Kraken's account, with what Kraken saw at each stage. Application: the applicant's email matched a list of hacker-linked addresses that industry partners had shared. Recruiter call: the applicant joined under a different name from the CV, then quickly changed it, and occasionally switched voices, a sign of real-time coaching. Background checks: the email led to a network of fake identities, the GitHub profile used an email exposed in a past breach, and the ID looked altered using details stolen two years earlier. Technical interviews, marked "kept in the process on purpose": Kraken kept going "not to hire, but to study their approach". Final interview with the security chief: confirm your location, hold your government ID up to the camera, and recommend local restaurants; the candidate failed. Below, Kraken's lesson: documents can be forged, but live, unprompted questions are hard to fake; genuine candidates pass them easily, so vary the questions. Two boxes add the Japanese picture: in May 2025 an applicant for an engineering job at a Japanese crypto exchange posed as someone else and was caught at interview; and North Korean IT workers used the AI face-swap tool Magicam for a few minutes, then advised each other to turn the camera off and blame the network.

Figure 4. Original graphic built from Kraken's account and the NPA's alert of 18 September 2026.

In a post on 1 May 2025, Kraken described an applicant for an engineering job who turned out to be North Korean. In the first call with a recruiter, the applicant joined under a different name from the one on the CV and quickly changed it, and occasionally switched voices, which suggested someone was coaching in real time. Industry partners had shared a list of email addresses linked to North Korean hackers, and the applicant's address was on it. Kraken's red team found that the address belonged to a network of fake identities, that the applicant's GitHub profile used an email exposed in an earlier data breach, and that the ID appeared to have been altered with details stolen in an identity theft two years before.

Kraken kept the applicant in the process, "not to hire, but to study their approach". In the final interview, its security chief asked the candidate to confirm their location, hold a government ID up to the camera and recommend restaurants in the city they claimed to live in. The candidate could not. Kraken's advice is that live, unprompted checks catch what documents miss, and that interviewers should vary the questions so they cannot be rehearsed.

The same thing in Japan

A Japanese exchange caught one too. The NPA's alert of 18 September says that in May 2025 an applicant for an engineering post at a Japanese crypto exchange posed as someone else, and the company noticed oddities in the interview and did not hire. The same alert describes IT workers using the AI face-swap tool Magicam for only a few minutes, then advising each other to turn the camera off and blame the network.

What to look for, according to Japan and its allies. On 31 July 2026, the NPA, the National Cybersecurity Office, the Ministry of Foreign Affairs, the Ministry of Finance and the Ministry of Economy, Trade and Industry issued a joint alert with the United States, South Korea and other partners (NPA; MOFA). North Korean IT workers pose as other nationals on job and freelance platforms, use translation services and language models to write convincing profiles, and can be spotted by mismatches with photo ID, video that looks altered or generated, refusal to join video calls, rates well below the market, signs that several people run one account, and requests to be paid in crypto. The alert asks platforms to check identity documents strictly and to require interviews in person.

The laptop farm

Flow diagram in two parts. The work: a North Korean IT worker abroad uses a borrowed or stolen identity and an AI-polished CV; your company hires a "remote engineer"; the laptop is shipped to the enabler's home, where a "laptop farm" holds many employers' laptops, which KVM switches and remote-access tools let the worker abroad operate, shown by a dashed "remote control" arrow back to the worker. The money: salary is paid into an account in the enabler's or a borrowed name, forwarded abroad, including as crypto, and becomes revenue for the state's weapons programmes. Two fact boxes. Japan, NPA, 18 September 2026: the first laptop farm for North Korean IT workers identified in Japan; several hundred million yen sent abroad, including crypto; Japanese enablers' home PCs, ID images and bank accounts were used. United States, Justice Department, 13 October 2026: 187 laptops and KVM devices seized from more than 40 homes in Texas; one Ohio indictment charging 14 people over a scheme affecting more than 70 U.S. companies and about $3 million; and the FBI in Dallas saying each laptop potentially represents one company that hired a North Korean worker. A final line notes that when exposed, one IT worker extorted a client and published its source code, and another defaced a client's website.

Figure 5. Original graphic built from the NPA's alert of 18 September 2026 and the U.S. Justice Department's actions of October 2026.

A fake remote employee needs a helper in the country. Companies ship laptops to a home address, and a login from abroad would stand out. So an "enabler" hosts the laptops at home and lets the worker abroad operate them through KVM switches or remote-access tools. In September 2026 the NPA said it had identified the first such laptop farm in Japan, with several hundred million yen, including crypto, sent abroad. Japanese enablers supplied PCs at home, their ID images were misused, and their bank accounts received the pay. The NPA warns that hiring or paying North Korean IT workers, or helping them earn foreign currency, may break Japanese law and sanctions.

In the United States, the Justice Department announced on 13 October 2026 that agents had searched more than 40 homes in Texas and seized 187 laptops and KVM devices (Northern District of Texas). The FBI in Dallas said each laptop potentially represents one company that hired a North Korean worker. A separate Ohio indictment charges 14 people over a scheme that reached more than 70 U.S. companies and earned about $3 million (Northern District of Ohio).

Exposure does not always end it. The NPA reports that one IT worker extorted a client and published its source code, and another defaced a client's website.

By the numbers

Two panels. Left, a bar chart of crypto thefts the FBI attributes to TraderTraitor, in US dollars at the time: DMM Bitcoin in May 2024, $308 million, way in a fake recruiter according to the FBI and the NPA; Bybit in February 2025, $1.5 billion, way in not covered in this article; Binance in October 2026, $6.34 billion, way in a tampered installer and a fake recruiter according to Binance, with a note that this is Binance's own estimate and the FBI says more than $6 billion. Right, five tiles: 30,000+ PCs in over 100 countries infected through fake job interviews between December 2025 and July 2026; 7,000+ crypto wallets emptied or credentials taken, with at least ¥1.7 billion moved; 57 suspected North Korea-linked cases Binance says it stopped in 2026; 187 laptops and KVM devices seized in Texas in September 2026; and 70+ U.S. companies in one Ohio indictment of fake remote workers.

Figure 6. Original graphic. Theft values are those the FBI or the victim gave at the time; Bybit's is from the FBI's alert of 26 February 2025.

Each theft is four to five times the last. Two of the three, DMM Bitcoin and Binance, ran through a fake recruiter. For the rest of this year's crypto losses, see our Q3 2026 review.

The attacker's playbook

The execution chain for each door, with MITRE ATT&CK techniques. Where a technique is our nearest match rather than the source's own label, we say so.

Door 1, the fake recruiter

  1. Choose the target from public profiles that name the employer and the role (T1593.001, Search Open Websites/Domains: Social Media). At Binance, a June compromise had already told the attacker who worked on wallets.
  2. Build a fake company and recruiter on legitimate job sites and LinkedIn (T1585.001, Establish Accounts: Social Media Accounts), complete with an "HR" interviewer, a "CEO" and personal social-media content.
  3. Make contact through the platform's messages (T1566.003, Spearphishing via Service).
  4. Get the target to run the code: an assignment repo, a pre-employment test or a "fix" for a video call (T1204.002, User Execution: Malicious File). A tampered installer found through a web search did the same job in Binance's June stage (nearest match: T1608.006, SEO Poisoning).
  5. Take the session, then the system: stolen session cookies at Ginco (T1539, Steal Web Session Cookie); deployment credentials and an internal AI tool at Binance (T1078, Valid Accounts).
  6. Move the money through signing systems and launder it quickly (T1657, Financial Theft).

Door 2, the fake applicant

  1. Borrow or steal an identity and write a convincing CV with language models (T1656, Impersonation).
  2. Get through interviews with help: a coach feeding answers, a face swap for a few minutes, then "network problems".
  3. Have the laptop sent to an enabler and run it from abroad through KVM switches or remote-access tools (T1219, Remote Access Tools).
  4. Work, get paid, and keep the access: salaries to an enabler's account, with insider access to code and data (T1078, Valid Accounts). When exposed, some extort (T1657, Financial Theft).

What does not change, whatever the next disguise: the recruiter needs the target to run code on a machine that matters, and the applicant needs the company to accept someone it has never verified live. Close those two conditions and the disguise stops mattering.

The defender's playbook

Two cards of red flags, each with a source in brackets. For the candidate, is the recruiter real: they found you through a profile that names your current employer and role (Binance); the company, the "Project Director" or the "CEO" cannot be verified independently (Binance); you are asked to clone a repo, run a "test", or install a tool to "fix" the video call (Binance, Ginco, WaterPlum); the task pulls in npm packages or a VS Code project you have never seen (WaterPlum); anything that would run on your work laptop (Binance); and a firm that says it is in AI, crypto or NFTs and moves very fast (WaterPlum). For the hirer, is the applicant real: the name on the call differs from the CV, or changes (Kraken); voices switch, someone else can be heard, or eyes keep going to another screen (Kraken, NPA); the camera goes off, the call keeps dropping "because of the network", or the face looks generated (NPA); they cannot answer simple live questions about the city they claim to live in (Kraken); email or GitHub is tied to breached data or other identities, or the ID looks altered (Kraken); and they want the laptop sent elsewhere, pay in crypto or to another person's account, or refuse to meet in person (NPA, Justice Department). A box below: never run a stranger's code on a work machine, and never let a document stand in for a live, unscripted conversation.

Figure 7. Original graphic. Every flag comes from a case or an advisory named in brackets.

Two chains of five stages, each with a control and the teams that own it. Door 1, the fake recruiter. Public profile: tell staff in sensitive roles that their profiles are a target list and not to name signing or wallet duties; owned by security and every employee. The approach: a simple, blame-free way to report an odd job offer; security and every employee. The test: job tasks and personal side work never on work devices, unknown code only in a throwaway virtual machine; every employee and IT. The payload: EDR on every developer machine, allow-lists for installers and packages, alerts on new npm installs from unknown repositories; IT and security. The pivot: signing systems unreachable from ordinary laptops, short-lived credentials, least privilege for internal AI tools, and two-person approval for large transfers; security and IT. Door 2, the fake applicant. Application: share and check lists of suspect emails and identities with industry peers; HR and security. Interviews: camera on throughout, live and unscripted local questions, in person for sensitive roles; HR and hiring managers. Offer and identity: check the ID live against the person, compare IP address and location, check the phone number is real; HR and security. The laptop: ship only to a verified home address, and flag remote-access tools, KVM devices and VPN logins from elsewhere; IT and security. Pay: only into an account in the employee's own name, never in crypto, and report suspicions to the police; finance and HR.

Figure 8. Original graphic. No single team can close either door alone.

For every employee

  • Treat any request to run code, install a tool or "fix" a meeting app during a job process as an attack. Do the tests, if at all, on a personal device you can wipe, never on a work machine.
  • Report odd job approaches. Make it blame-free: an engineer looking for a new job is normal, and the report is what lets security warn colleagues.
  • Keep sensitive duties off public profiles. Binance's attacker learned who worked on wallets from its June foothold; others learn it from profiles.

For HR and hiring managers

  • Keep the camera on for the whole interview, and ask live questions that need no preparation: the neighbourhood, the weather, the nearest station. Vary them, as Kraken advises.
  • Check identity live: the ID against the face on camera, the phone number, and whether the applicant's internet address matches where they say they live, as the NPA recommends.
  • Meet in person for roles near money, keys or production systems. The 31 July alert asks platforms to require it; companies can too.
  • Share intelligence. Kraken was tipped off by peers' lists of suspect emails. Join the industry channels you have, and contribute to them.

For IT and security

  • Make developer machines boring targets: EDR everywhere, allow-lists for installers and packages, and no admin rights by default. Binance's June stage began with an installer found through a web search.
  • Keep signing and production systems out of reach of ordinary laptops. Use separate hardened devices, short-lived credentials and two-person approval for large transfers. Give internal AI tools only the privileges their users need, because at Binance one became a stepping stone.
  • Watch new hires' devices: remote-access tools, KVM-over-IP devices, and VPN logins from places the employee does not live.

For finance and legal

  • Pay only into accounts in the employee's own name, in the currency of their contract, never in crypto.
  • Know the legal exposure. The NPA warns that paying North Korean IT workers can breach Japanese law and sanctions, even unknowingly; report suspicions to the police promptly.

Copy-paste templates

Note to all staff

Subject: Job offers that ask you to run code

North Korean operatives are posing as recruiters on LinkedIn and job sites.
At Binance, an "interview assignment" run on a work computer was one step
in an attack that took about $6.3 billion in October 2026.

- Never run a coding test, repo, installer or "meeting fix" from a recruiter
  on a work device. If you must do a test, use a personal device you can wipe.
- If a recruiter, "HR" contact or "CEO" asks you to install or run anything,
  stop and report it to [security contact]. You will not be in trouble for
  looking for a job.
- Do not list wallet, signing or production duties on public profiles.

Thank you.

Interview identity checks for hiring managers

INTERVIEW IDENTITY CHECKS – REMOTE ROLES
Candidate:                     Role:                    Interviewer / date:

[ ] Camera on for the whole interview; name on screen matches the CV
[ ] Only one voice; no signs of coaching or a second screen
[ ] Two or three live, unscripted local questions (vary them each time)
[ ] Government ID shown on camera and matched to the face
[ ] Phone number called and answered by the candidate
[ ] Connection location consistent with stated home (security to check)
[ ] Email / GitHub checked against shared lists of suspect identities
[ ] In-person meeting arranged for roles near money, keys or production
[ ] Laptop to a verified home address only; pay to an account in their own name
[ ] Anything odd reported to security before an offer is made

What remains unknown

  • How Binance's stages connect. Binance's first update pointed to one DevOps laptop; later updates found more points of breach. It says the investigation continues and will be reviewed independently.
  • Which campaign reached Engineer B. Binance describes the fake interview but has not tied it to a named campaign.
  • Who Kraken's applicant was, and whether the same identity was hired elsewhere.
  • How many Japanese companies have unknowingly hired North Korean IT workers. The NPA has described one laptop farm; it has not given a count of employers.
  • Who the Japanese exchange was. The NPA does not name it.

Sources and updates

Published 4 November 2026 (JST). Our research network blocks direct access to most of these sites. We therefore read Binance's updates on X, Kraken's blog post, the FBI's alerts, the NPA's and MOFA's pages and the Justice Department's releases through search-engine extracts of those pages, and we quote only what those extracts showed. The WaterPlum figures also match the joint advisory as we read it directly for our earlier analysis of the four state actors. Binance's $6.34 billion is its own estimate; the FBI says "more than $6 billion". Binance's 27 October statement is cited through Wu Blockchain's relay because we could not find Binance's own post.

All articles →日本語で読む ↗