Colourful shipping containers stacked high at a port
Cybersecurity

Software Supply Chain Attacks: When a Trusted Package Turns

One phished developer, billions of weekly downloads. How the 2025 npm supply chain attacks unfolded, why they work, and how to protect your apps and your team.

F

Force Media Team

7 min read

Photo: Julia Taubitz on Unsplash

On 8 September 2025, a developer called Josh Junon opened an email that looked like it came from npm, the registry where JavaScript developers publish and download code. It said his account would be locked unless he updated his two-factor authentication within two days. The link went to npmjs.help, a domain registered three days earlier.

He entered his details. Soon afterwards, new versions of packages he maintains, including chalk and debug, had been published with malicious code inside. According to Check Point's write-up, the 18-plus packages affected had over 2 billion weekly downloads between them.

Nobody hacked a bank or a government. They hacked one person, and through him, a huge slice of the world's software. That's a software supply chain attack, and 2025 was the year it stopped being a niche concern.

What a software supply chain attack actually is

Almost no modern software is written from scratch. A typical web application pulls in hundreds of open-source packages: small, free building blocks for things like formatting dates, colouring terminal output or talking to a database. Those packages pull in their own packages, and so on. Your app ends up trusting code written by thousands of people you've never heard of.

A supply chain attack poisons one of those building blocks, so that everyone who installs it gets the poison too. The attacker doesn't need to break into your systems. You invite the code in yourself, during a routine install or update.

How the September 2025 npm compromise worked

  1. 01

    Phish

    A fake 'npm support' email sends the maintainer to a lookalike login page.

  2. 02

    Take over

    With his credentials and 2FA code, the attackers control his npm account.

  3. 03

    Publish

    Malicious versions of trusted packages go up on the official registry.

  4. 04

    Spread

    Developers and build systems around the world install the update automatically.

  5. 05

    Steal

    In browsers, the code quietly swaps cryptocurrency wallet addresses for the attackers' own.

The compromise was spotted and made public the same day, and npm began pulling the malicious versions that evening. Anything built or deployed in that window could have shipped the payload. Source: Check Point Research.

Then came the worm

A week later, something worse appeared. A self-spreading piece of malware, nicknamed Shai-Hulud, started moving through npm on its own. The US Cybersecurity and Infrastructure Security Agency issued an alert on 23 September 2025 saying it had compromised over 500 packages.

Here's what made it different. When a developer installed an infected package, the worm searched their machine for secrets: GitHub access tokens and keys for AWS, Google Cloud and Microsoft Azure. It sent them to the attackers. Then it used the developer's own npm access to inject itself into their packages and publish new infected versions. Every victim became a new carrier.

In late November 2025 a second wave arrived. Check Point's analysis of Shai-Hulud 2.0 counted:

621

npm packages infected in the November wave

Check Point, 2025

25,000

GitHub repositories compromised

Check Point, 2025

14,206

secrets leaked, of which 2,485 were still valid

Check Point, 2025

The second version ran its code during the install step itself, before any of the package was even used, and carried a destructive fallback that could wipe files if it detected it was being analysed.

Green terminal text scrolling on a dark screen during a software package install
A routine install scrolls past hundreds of packages in seconds. Nobody reads them all, which is exactly what attackers count on. Photo: Jake Walker on Unsplash

This isn't a one-off

The headline incidents sit on top of a much larger, quieter problem: packages published as malware from day one, with names designed to be mistaken for popular ones. Sonatype, which tracks open-source malware, reports a steep climb:

Known malicious open-source packages, cumulative

Sonatype identified more than 454,600 new malicious packages in 2025 alone, and says over 99% of the year's open-source malware was on npm. Source: Sonatype, State of the Software Supply Chain 2024 and 2026

npm is the biggest target because it's the biggest registry, not because it's uniquely careless. Python's PyPI and other ecosystems see the same tactics at a smaller scale.

You don't have to be the target. You just have to install the same package as everyone else.

Why this matters to a business that isn't a tech company

If you've had a website, app or internal system built in the last few years, it almost certainly depends on npm packages, even if nobody on your team has ever heard of them. The risks are concrete:

  • Your customers' data or money. The September payload targeted cryptocurrency transactions in the browser; a different payload could just as easily skim form data.
  • Your cloud bill and accounts. Stolen cloud keys can be used to run up huge bills, or to get into your hosting and databases.
  • Your reputation. If your site serves malicious code to visitors, it's your name on it, whoever wrote the package.

It's the same lesson as the WordPress plugin problem and the React2Shell vulnerability: the code you didn't write is part of your attack surface.

How to protect your projects

CISA's guidance after Shai-Hulud is a good backbone. This is how we'd put it into practice, in order:

  1. 1

    Know what you're running

    Make sure every project has a lock file (package-lock.json or yarn.lock) committed, so the exact version of every package, including nested ones, is recorded and reviewable.

  2. 2

    Don't take updates blindly

    Pin dependencies to known-good versions and update deliberately, reviewing what changed. Some teams now wait a few days before adopting a brand-new release, because most malicious versions are caught quickly.

  3. 3

    Lock down developer accounts

    Require phishing-resistant multi-factor authentication (passkeys or security keys) on GitHub, npm and cloud accounts. A code sent by text or app can be phished; a hardware key bound to the real site can't.

  4. 4

    Keep secrets off developer machines

    Use short-lived tokens and a proper secrets manager. The fewer long-lived keys sitting in files and environment variables, the less a worm can steal.

  5. 5

    Turn on automated alerts

    GitHub's Dependabot alerts and secret scanning, or an equivalent tool, warn you when a package you use is flagged or a key is exposed.

  6. 6

    Have a plan for 'it happened'

    Know how you'd find which projects use an affected package, and how you'd rotate every credential quickly. CISA's first advice after Shai-Hulud was to rotate all developer credentials immediately.

If you're not sure what your website or app depends on, or who's watching for these alerts, that's worth finding out before the next incident rather than during it. Our ransomware survival guide covers the wider basics of keeping a small business safe.


Force Media builds and maintains web applications for businesses in Sri Lanka, Australia and beyond, with dependency monitoring, lock-file hygiene and secure deployment built in. Ask us to review your project's dependencies, and we'll tell you what you're trusting and whether you should be.

Speak to our experts

From AI agents to secure, fast websites, we build technology that does real work for your business.

Get in touch
FAQ

Common Questions

Everything you need to know about working with Force Media.

Ask Us Anything
We primarily serve businesses in Sri Lanka and Australia, but our remote-first approach means we can work with clients globally.
Project timelines vary by scope. A standard website takes 2–4 weeks. Custom software projects range from 6–16 weeks depending on complexity.
Yes. All our projects include a support period, and we offer ongoing retainer packages for maintenance, SEO, and feature development.
Absolutely. We specialize in integrating with existing platforms, APIs, and legacy systems to extend their capabilities.