theSIGNAL TECHNOLOGY
4 November 2026
"Amateurs hack systems; professionals hack people."
Photo: Mika Baumeister / Unsplash
A developer works at a laptop, the kind of everyday target attackers study for months.
🛡️
57
suspected North Korea-linked cyber threat cases Binance says it has pr
Discuss
  • Why might attackers choose months of patient social engineering over a single technical exploit against a well-defended company?
  • Binance's security tools caught the first infection but missed the second. What does this suggest about relying on automated detection?
Technology

How a Fake Job Interview Cracked Crypto's Vault

Binance's postmortem of the largest crypto theft on record shows patient, human-centred deception beating security tools built to stop it.

The most expensive crypto heist in history did not begin with a brilliant exploit. It began with a Google ad. According to a postmortem published by Binance, one of its developers downloaded a tampered installer in June, apparently spread through a malicious advertisement, and in doing so exposed details about the company's engineers and how they work. What followed was a slow campaign of trust-building.

In September, a different engineer was approached by someone posing as a recruiter. There were calls, multiple rounds of interviews and shared social media profiles, all designed to look ordinary. Finally came a "take-home assignment", which was in fact malware. It deleted a legitimate development tool, pyenv, and replaced it with a malicious copy that gave the attackers a backdoor.

Binance says its tools missed this second attack, even though its security teams had caught the first infection in June and replaced the worker's device. The FBI attributes the operation to a North Korean unit called "TraderTraitor". By October 20th, Binance says, the group had used the company's own internal AI tool to gain control over wallets linked to the first compromised engineer and to a third employee. On October 29th it logged into the third employee's computer and struck.

//
Because the hot wallet used Multi-Party Computation (MPC), there was no single private key that could be used to immediately revoke the malicious EIP-7702 authorizations.
Binance security postmortem
Technology

Can You Prove Which AI Answered You?

When you pay for a premium AI model, how do you know the answer really came from it? Anthropic's researchers have published FlashProve, an open source system that attaches a cryptographic guarantee to an output, much like a digital signature on an email. Labs would publish "fingerprints" of their models, which reveal no private weights, so even a lab could not secretly swap in a cheaper model.…
  • Would you trust AI companies more if every answer carried verifiable proof?
Technology

A Fake Job Interview Cost $6.3 Billion

Two interviews and one "coding test" were enough. Binance says North Korean hackers, posing as recruiters, infected a wallet engineer's laptop, then pivoted to two colleagues, one already compromised by a tampered download months earlier. The result was a record $6.3 billion theft. The weakest link was not code but courtesy: a hiring process people trust.…
  • If hiring itself can be weaponized, how should companies verify recruiters and protect their employees?
AMERICAS · Technology
I read Anthropic's 92-page risk factor document, so you don't have to – video explainer
ASIA · Technology
China publicly discloses plan for a lunar landing within three years
GLOBAL · Technology
These firms you've never heard of are feeding fears the AI bubble could burst
EUROPE · Technology
Ukraine’s new pocket-sized gadget turns Russian drones’ cameras against them
theSIGNAL IN THE LAB
1VOCABULARY
tamperedmalicious
backdoorcompromisedpivoted
revokesafeguard
2GRAMMAR FOCUS
Participle clauses — present (-ing) and past (-ed)
A present participle (-ing) clause gives an active meaning and shares its subject with the main clause, while a past participle (-ed) clause gives a passive meaning. Both can replace a longer clause to add information about time, reason or result.
posing · spread · replacing · exposing · linked · gaining · caught
  1. The developer downloaded a tampered installer, apparently through a malicious advertisement.
  2. A different engineer was approached by someone as a recruiter, and the conversation seemed completely ordinary.
  3. The first infection was dealt with quickly, with security teams the worker's device in June.
  4. The attackers installed malware on the laptop, them a backdoor into the company's systems.
  5. in June, the first infection should have warned Binance that its engineers were being targeted.
  6. The hackers used the company's internal AI tool, control over wallets linked to two employees.
3DEPENDENT PREPOSITIONS
Complete each phrase with the correct preposition. All from today’s articles.
  1. The FBI attributes the operation ___ a North Korean unit called "TraderTraitor".
  2. The developer's download exposed details ___ the company's engineers.
  3. An engineer was approached by someone posing ___ a recruiter.
  4. The group used the company's AI tool to gain control ___ several wallets.
  5. The malicious copy was a replacement ___ a legitimate development tool.
4CRITICAL THINKING
The articles suggest that the human element, not the technology, was the real point of failure. Who bears more responsibility when an employee falls for a sophisticated scam: the individual who clicked, or the company that did not prepare them? Consider how far a company can reasonably go in policing what staff do on personal devices and social media.
5CREATIVE · HEADLINES
Write a headline for the top story in each of the following styles. One line each, no explanation:
  • TABLOID NEWSPAPER
  • LUXURY MAGAZINE
  • ACTIVIST BLOG
6WRITING
Imagine you are a security officer at a mid-sized tech company. Write a short internal email (about 100 words) to all staff explaining how to respond if a recruiter contacts them with a take-home assignment, using lessons from the Binance case.
7DEGREES OF EXTREMITY
Complete each ladder from mild to strong.
  • a worrying breach→→
  • slightly suspicious→→
  • a minor setback→→
  • quite careless→→
  • fairly convincing→→
  • a little uncomfortable→→
8SPEAKING
  1. Have you ever nearly fallen for a scam? What gave it away?
  2. Which is worse for a company: losing money or losing public trust?
  3. Is it ever acceptable for hackers to target the vulnerable, such as job-seekers?
  4. What personal data would you refuse to share with any employer, even a trusted one?