PushWarden began as a hunter for PolinRider. Other campaigns attack developers the same way, so it checks for the tricks they share too, and its general rules catch many new variants.
PolinRider
Critical danger
Hidden in files you open every day
- How it reaches you
- Code added to config files such as
postcss.config.mjs or an app's entry file, a script disguised as a font, or an editor task that runs the moment you open the folder.
- What it does
- A loader fetches more code from the attackers' servers, takes tokens and keys (GitHub, npm, cloud), sets up remote access that survives a restart, and uses your GitHub token to push itself into every repository it can reach. Your colleagues who pull become the next victims. Researchers connect it to the Contagious Interview operation.
- What PushWarden does
- Finds the infected file and quarantines it; with Pro only the malicious part is cut out. Stops the running process, removes its start-up entries and, with Pro, blocks its servers.
github-clean finds every infected branch and fixes it with a normal commit.
Read more: Socket · OpenSourceMalware dossier
Shai-Hulud
Critical danger
A worm that spreads through npm
- How it reaches you
- An install script inside an npm package runs when you install it.
- What it does
- Steals the tokens it finds and uses them to publish copies of itself inside other packages, so one infection can reach everyone who installs those packages.
- What PushWarden does
- Reports the infected package files and never moves
package.json by itself, so your project is not broken without your say-so.
Read more: Unit 42
Code you cannot see
- How it reaches you
- Malicious code written in invisible characters, so it looks like an empty line in your editor and in a code review.
- What it does
- Runs hidden code with your permissions. Because nothing shows on screen, it can sit in a project unnoticed.
- What PushWarden does
- Quarantines the file and shows what was hidden in it.
Read more: Veracode
Your own AI assistant used against you
- How it reaches you
- An install script in a compromised package (the Nx case) starts your AI coding tool with its safety checks switched off.
- What it does
- Uses your own tool to look through your files for secrets and sends them out, so a normal-looking command does the stealing.
- What PushWarden does
- Reports the script and the AI-tool settings it relies on, so you can remove them and rotate what was exposed.
Contagious Interview
High danger
A fake job, a real spy
- How it reaches you
- A recruiter sends a "coding test". Running or opening the project starts the malware.
- What it does
- Installs a keylogger and a screen grabber. It watches what you type and see, which can include passwords, wallet seed phrases and company access.
- What PushWarden does
- Reports the project's malicious files and hooks, so you can delete the folder and change what you typed while it ran.
Mini Shai-Hulud and AI tool hooks
High danger
Runs when you open the folder
- How it reaches you
- A task or hook, in the editor or in an AI tool's settings, that runs a hidden script as soon as the folder is opened.
- What it does
- Downloads and runs more code with no click from you, which is why opening an unknown project is risky.
- What PushWarden does
- Quarantines auto-run tasks that download code and reports the rest.
Read more: OpenSourceMalware
Git and CI droppers
Medium danger
Quiet changes to git and pipelines
- How it reaches you
- Git settings and hooks, dev containers or workflow files that run a command, or a GitHub Actions workflow that sends secrets out.
- What it does
- Runs a command on every commit, clone or build, and can leak the secrets your pipeline holds. A leaking workflow can be as serious as the ones above.
- What PushWarden does
- Quarantines hooks and tasks that download code and reports the rest, with the reason for each finding.
PushWarden