Amazon Ties North Korean Hacking Group to Four Major Open-Source Software Compromises

Amazon says a North Korea-linked hacking group has compromised four major open-source JavaScript packages since March 2025, using stolen trust rather than brute-force hacking to slip malicious code into software used by companies worldwide.

The assessment, published Wednesday in a blog post by Amazon Integrated Security CISO CJ Moses, links the group for the first time to compromises of four packages: typo-crypto, debug, chalk and axios. According to Amazon, the axios package alone gets more than 100 million downloads every week. Its compromise had already been attributed to North Korean hackers before this new report tied it to the same broader campaign as the other three.

Amazon Threat Intelligence attributed the campaign to the group with "medium confidence," citing reused code and overlapping attack techniques across the four incidents, according to the blog post. Researchers track the group under multiple names: Sapphire Sleet, Stardust Chollima, BlueNoroff, CageyChameleon and Alluring Pisces. That alphabet soup of aliases reflects how different cybersecurity firms have independently discovered and named the same actor over time.

How the scheme worked

In each case, according to Amazon, the hackers tricked a trusted software maintainer into granting them access, then used that access to publish a software update laced with malicious code. Organizations that automatically pull the latest version of a package, standard practice for a lot of development shops, would have downloaded the compromised code straight into their own systems without ever knowing it.

Instead of breaking into thousands of companies one at a time, attackers can compromise one widely used building block and let those companies infect themselves through routine software updates.

Why open source is a soft target

Open-source code, the freely available software that underpins operating systems, web servers, encryption tools and countless business applications, runs on a trust system. Volunteer maintainers review changes, patch bugs and push updates, often without pay and without the security vetting a corporate software vendor would apply.

Attackers can spend weeks or months posing as helpful contributors, fixing minor bugs and building credibility, before attempting to seize control of a project or sneak a backdoor into an update.

Amazon's report echoes a dynamic that got broad attention in 2024, when an account using the name "Jia Tan" spent years cultivating trust among developers before trying to plant a backdoor in XZ Utils, a compression tool built into numerous Linux distributions. That backdoor was caught before it could spread widely. This time, according to Amazon's findings, the North Korean-linked group appears to have succeeded in getting malicious updates published across four separate packages before being detected.

The volunteer-maintainer problem

The open-source community's reliance on unpaid maintainers is the mechanism attackers are using. If a project has one or two maintainers doing this in their spare time, it's a lot easier to earn their trust, or exhaust their vigilance, than it is to breach a company with a dedicated security team.

Critics of the open-source trust model have pushed for years for better-funded maintainer programs, mandatory code-signing, and stricter vetting before someone gets commit access to a package with 100 million weekly downloads. Amazon's findings give that argument fresh weight, since the axios compromise shows just how much downstream exposure a single infiltrated maintainer account can create.

At the same time, open-source defenders have a fair point: the same transparency that makes these packages exploitable is also what allowed Amazon's researchers, and others before them, to catch the malicious code, trace it, and attribute it. A closed-source proprietary system compromised the same way might never have been caught at all.

No companies that downloaded the compromised packages have been named publicly in Amazon's disclosure, and it's not yet clear how many organizations pulled the malicious updates before they were flagged and removed. Amazon has not said whether it has notified specific affected customers or referred the findings to U.S. law enforcement.

What happens next depends largely on the open-source maintainers themselves. Package registries like npm, which hosts the JavaScript ecosystem where all four compromised packages live, have faced pressure to tighten publishing requirements, including mandatory two-factor authentication for maintainers with high-download packages. Whether that gets enforced more broadly following Amazon's report is the open question security researchers will be watching.