The job interview was the attack: fake recruiters are the new phishing

Binance has now explained how attackers stole $6.43 billion from it, and the story starts with a job interview. A friendly recruiter, two rounds of video calls, and a coding "assignment" at the end. The engineer ran it on his work computer, and that was the way in. This isn't a crypto problem. It's a phishing problem dressed up as a career opportunity, and it can land on anyone who writes scripts, manages servers or holds admin rights.

What happened at Binance

According to Binance's account, reported by The Hacker News, there were two separate ways in. Back in June, one engineer searched for an AI tool, found a fake download site, and installed a tampered Homebrew installer. Then in September, someone posing as a "Project Director" contacted a second engineer through a real job-hunting platform.

The fake hiring process was convincing. The first round was a video interview with an "HR representative", who even shared a social media account full of personal photos and videos to build trust. The second round was with someone claiming to be the CEO of a company that didn't exist. Near the end, the engineer was asked to download an assignment, a code repository, onto his work computer. Setting it up ran malicious code.

From there the attacker moved carefully. The malware swapped a normal development tool component for a poisoned one, talked to the attacker's server, and cleaned up after itself. In October the attacker used that computer and abused an internal AI tool to get at other engineers' machines, found a deployment token, and finally used an internal wallet signing service to send transactions outside the normal process.

Why fake interviews work so well

Most phishing training teaches people to spot a bad email. A fake job offer gets around almost all of that:

It arrives on a trusted platform. A message on a real job site doesn't look like spam.
There are real people on video. Many of us were taught that a video call proves someone is genuine. Here it was part of the trick.
Running the code is the whole point. In a coding test, downloading and running a stranger's project is what you're expected to do. There's no "suspicious attachment" moment.
People keep it quiet. Hardly anyone tells their manager they're interviewing elsewhere, so nobody else gets a chance to say "that looks odd".

This isn't new either. Gen Digital wrote up a very similar case in February 2025: a recruiter on LinkedIn, a front-end project in a Bitbucket repository, and a hidden file that stole saved passwords and crypto wallet data, then installed a remote access tool and a cryptominer.

What I'd tell staff

You can't stop people looking for jobs, and you shouldn't try. But you can make the rules clear and blame-free:

Never run a recruiter's code on a work device. Not a test, not a "quick setup script", not an installer. If a test needs code, use a personal machine or, better still, a throwaway virtual machine that holds no work logins.
Check the company exists outside the conversation. Look it up yourself, and don't rely on links or profiles the recruiter sends.
Treat pressure as a warning sign. Offers that are too generous, recruiters who are too keen, and deadlines measured in hours are all red flags.
Report it without fear. Say plainly that nobody gets in trouble for reporting a dodgy recruiter, even if they were job hunting.
Get AI tools from the official site. The first way into Binance was a fake download of an AI tool found through a search engine.

What I'd check as the engineer

The interview was only the start. The damage came from what that one computer could reach. A few practical checks:

Tokens and secrets on laptops. The attacker found a deployment token sitting on an engineer's machine. Look for long-lived tokens, saved credentials and SSH keys on admin and developer devices, and move them into a proper vault with short expiry times.
Separate admin from everyday work. The device someone uses for email and browsing shouldn't also hold the keys to production. Separate admin accounts, and where you can, separate admin workstations.
Internal AI tools are part of your attack surface. At Binance an internal AI tool helped the attacker reach other machines. Anything that can act on your systems needs the same access controls and logging as a human admin.
Block unknown code from running. Application control, SmartScreen and good endpoint protection won't catch everything, but they raise the bar for a random script from a stranger.

The short version

A job offer is now one of the best disguises an attacker has. The fix isn't to spot a better fake, because these ones are very good. It's a simple rule that holds whatever the story: work devices don't run code from strangers. Back that with fewer secrets lying around on laptops, and a single bad interview stays a bad interview instead of a breach.

Sources

← All posts