Four distinct compromises of open source Node Package Manager (NPM) libraries can now been confidently linked to one threat actor backed by the North Korean government, says Amazon’s threat research team.
At a media roundtable event at its Arlington, Virginia headquarters this week, Amazon shared evidence that the compromises of the axios, chalk, debug and typo-crypto libraries were all orchestrated by an advanced persistent threat (APT) actor tracked as APT38, Sapphire Sleet, or Stardust Chollima among other names, which likely operates under the more well-known Lazarus Group umbrella, linked to Pyongyang’s Reconnaissance General Bureau (RGB).
The axios compromise of March 2026 saw two new NPM packages for the widely-used HTTP client for JavaScript, injected with a malicious dependency that connected to command-and-control (C2) infrastructure and thence deployed a second-stage remote access trojan (RAT).
The hit on the axios package had already been linked to this particular threat actor, but by working backwards and examining the group’s operational playbook, Amazon now asserts that the group was also behind the March 2025 compromise of typo-crypto, and the chalk and debug compromises of September 2025. It believes the first attack on the less well-known typo-crypto package to have been a rehearsal run enabling the group’s operatives to test their playbook without making too much noise.
“We found one hand on the keyboard behind four different supply chain attacks,” CJ Moses, CISO of Amazon Integrated Security, told reporters. “When you connect these dots, this stops being a series of isolated incidents and starts looking more like an industrial operation, in this case perpetrated by a nation-state actor.”
In all four instances, said Moses, the group used social engineering techniques to gain the confidence of a trusted maintainer of the package, before pushing a malicious software update that went to any organisation unfortunate enough to pull the tainted version. This modus operandi is similar to that used against the XZ Utils data compression library back in 2024, in which threat actors spent years gaining the trust of the project maintainers with helpful contributions before attempting to sneak in a backdoor vulnerability that could have had serious consequences had a researcher not spotted it.
“Why does this matter? With open source software being a shared infrastructure, one compromised package can cascade to thousands of organisations in an instant,” said Moses.
“If you think back a couple of years to the XZ Utils backdoor, it demonstrated how you can have a patient attacker that is able to compromise critical open source packages, exploiting the trust placed in them.
“When we say the axios package has more than 100 million weekly downloads, that number represents real organisations pulling that code into production systems every single week,” said Moses, who noted that according to Wiz Research, one in 10 cloud environments were affected by the chalk and debug compromises in just two hours.
“The distribution mechanism that makes open source software so enticing and highly used is the same thing that these adversaries are using to their advantage as well. So for the DPRK [Democratic People’s Republic of Korea] this is a financially motivated pattern … compromising a small number of highly popular packaged gains potential access to thousands of downstream environments almost simultaneously.
“They’re not breaking the open source systems,” he added, “they’re using it as designed, just for a different purpose.”
Indeed, Moses said that since the XZ Utils incident, Amazon has seen the volume and sophistication of such software supply chain attacks increase, driven in large part by the North Koreans and threat actors and cyber criminal groups associated with them.
Amazon’s response
Recognising the increasing danger arising from open source supply chain attacks, Amazon said it was investing across its Threat Intelligence and Inspector teams to help its customers adapt to the shifting landscape.
It is working alongside industry partners such as the Open Source Security Foundation (OpenSSF) to share its findings, and earlier this year signed up alongside the Linux Foundation to launch Akrites, a collaboration that aims to defend open source software from artificial intelligence (AI) enabled threats. Amazon has ploughed over $12m into this initiative, saying the security of open source software is a shared responsibility.
