> Both versions were published using the compromised npm credentials of a lead axios maintainer, bypassing the project's normal GitHub Actions CI/CD pipeline.
Doesn’t npm mandate 2FA as of some time last year? How was that bypassed?
How much do you want to bet me that the credential was stolen during the previous LiteLLM incident? At what point are we going to have to stop using these package managers because it's not secure? I've got to admit, it's got me nervous to use Python or Node.js these days, but it's really a universal problem.
I can't even imagine the scale of the impact with Axios being compromised, nearly every other project uses it for some reason instead of fetch (I never understood why).
Also from the report:
> Neither malicious version contains a single line of malicious code inside axios itself. Instead, both inject a fake dependency, plain-crypto-js@4.2.1, a package that is never imported anywhere in the axios source, whose only purpose is to run a postinstall script that deploys a cross-platform remote access trojan (RAT)
Good news for pnpm/bun users who have to manually approve postinstall scripts.
Has anyone tested general purpose malware detection on supply chains ? Like clamscan . I tried to test the LiteLLM hack but the affected packages had been pulled. Windows Defender AV has an inference based detector that may work when signatures have not yet been published
PSA: npm/bun/pnpm/uv now all support setting a minimum release age for packages.
I also have `ignore-scripts=true` in my ~/.npmrc. Based on the analysis, that alone would have mitigated the vulnerability. bun and pnpm do not execute lifecycle scripts by default.
Here's how to set global configs to set min release age to 7 days:
(Side note, it's wild that npm, bun, and pnpm have all decided to use different time units for this configuration.)
If you're developing with LLM agents, you should also update your AGENTS.md/CLAUDE.md file with some guidance on how to handle failures stemming from this config as they will cause the agent to unproductively spin its wheels.
I recommend everyone to use bwrap if you're on linux and alias all package managers / anything that has post build logic with it.
I have bwrap configured to override: npm, pip, cargo, mvn, gradle, everything you can think of and I only give it the access it needs, strip anything that is useless to it anyway, deny dbus, sockets, everything. SSH is forwarded via socket (ssh-add).
This limits the blast radius to your CWD and package manager caches and often won't even work since the malware usually expects some things to be available which are not in a permissionless sandbox.
You can think of it as running a docker container, but without the requirement of having to have an image. It is the same thing flatpak is based on.
As for server deployments, container hardening is your friend. Most supply chain attacks target build scripts so as long as you treat your CI/CD as an untrusted environment you should be good - there's quite a few resources on this so won't go into detail.
This may not be popular, but is there a place for required human actions or just timed actions to slow down things like this? For instance, maybe a GH action to deploy requires a final human click and to change that to cli has a 3 day cooling period with mandatory security emails sent out. Similarly, you switch to read only for 6 hrs after an email change. There are holes in these ideas but the basic concept is to treat security more like physical security, your goal isn't always to 100% block but instead to slow an attacker for xxx minutes to give the rest of the team time to figure out what is going on.
Min release age sucks, but we’ve been here before. Email attachments used to just run wild too, then everyone added quarantine delays and file blocking and other frictions... and it eventually kinda/sorta worked. This does feel worse, though, with fewer chokepoints and execution as a natural part of the expectation.
Edit: bottom line is installs are gonna get SOOO much more complicated. You can already see the solution surface... Cooling periods, maintainer profiling, sandbox detonation, lockfile diffing, weird publish path checks. All adds up to one giant PITA for fast easy dev.
Genuinely how are you supposed to make sure that none of the software you have on your system pulls this in?
It’s things like this that make me want to swap to Qubes permanently, simply as to not have my password manager in the same context as compiling software ever.
Supply chain attacks are so scary that I think most companies are going to use agents to hard fork their own versions of a lot of these core libraries instead. It wasn’t practical before. It’s definitely much more doable today.
Once you fork it, you are then on the hook for forking every future update and security patch. You can automate a lot of the testing, but its still adding an extra failure point that you are now responsible for.
And if you pin it, then when you inevitably get a CVE for an old version, the upgrade path is harder and more time consuming. And that's when the security teams come knocking to pass their audit.
There are so many scanners these days these things get caught pretty quick. I think we need either npm or someone else to have a registry that only lets through packages that pass these scanners. Can even do the virustotal thing of aggregating reports by multiple scanners. NPM publishes attestation for trusted build environments. Google has oss-rebuild.
All it takes is an `npm config set` to switch registries anyways. The hard part is having a central party that is able to convince all the various security companies to collaborate rather than having dozens of different registries each from each company.
Rather than just a hard-coded delay, I think having policies on what checks must pass first makes sense with overrides for when CVEs show up.
Not to beat a dead horse but I see this again and again with dependencies. Each time I get more worried that the same will happen with rust. I understand the fat std library approach won’t work but I really still want a good solution where I can trust packages to be safe and high quality.
This is why Node.js is completely unsuitable as backend.
Until recently, there wasn’t even a standard Promise-based HTTP client. Why should we need to download a library just to make a simple HTTP request? It’s because Node.js’s standard library is too limited, leading to an explosive growth in third-party libraries. As a result, it’s vulnerable to security attacks, and maintaining it in an enterprise environment becomes a major challenge.
Let’s use .NET or Go. Why use JavaScript outside of the browser when there are excellent backend environments out there?
175 comments
[ 0.20 ms ] story [ 85.8 ms ] threadDoesn’t npm mandate 2FA as of some time last year? How was that bypassed?
https://github.com/axios/axios/issues/10604
Also from the report:
> Neither malicious version contains a single line of malicious code inside axios itself. Instead, both inject a fake dependency, plain-crypto-js@4.2.1, a package that is never imported anywhere in the axios source, whose only purpose is to run a postinstall script that deploys a cross-platform remote access trojan (RAT)
Good news for pnpm/bun users who have to manually approve postinstall scripts.
I also have `ignore-scripts=true` in my ~/.npmrc. Based on the analysis, that alone would have mitigated the vulnerability. bun and pnpm do not execute lifecycle scripts by default.
Here's how to set global configs to set min release age to 7 days:
(Side note, it's wild that npm, bun, and pnpm have all decided to use different time units for this configuration.)If you're developing with LLM agents, you should also update your AGENTS.md/CLAUDE.md file with some guidance on how to handle failures stemming from this config as they will cause the agent to unproductively spin its wheels.
I have bwrap configured to override: npm, pip, cargo, mvn, gradle, everything you can think of and I only give it the access it needs, strip anything that is useless to it anyway, deny dbus, sockets, everything. SSH is forwarded via socket (ssh-add).
This limits the blast radius to your CWD and package manager caches and often won't even work since the malware usually expects some things to be available which are not in a permissionless sandbox.
You can think of it as running a docker container, but without the requirement of having to have an image. It is the same thing flatpak is based on.
As for server deployments, container hardening is your friend. Most supply chain attacks target build scripts so as long as you treat your CI/CD as an untrusted environment you should be good - there's quite a few resources on this so won't go into detail.
Bonus points: use the same sandbox for AI.
Stay safe out there.
Edit: bottom line is installs are gonna get SOOO much more complicated. You can already see the solution surface... Cooling periods, maintainer profiling, sandbox detonation, lockfile diffing, weird publish path checks. All adds up to one giant PITA for fast easy dev.
It won't stop all attacks but definitely would stop some of these
It’s things like this that make me want to swap to Qubes permanently, simply as to not have my password manager in the same context as compiling software ever.
And if you pin it, then when you inevitably get a CVE for an old version, the upgrade path is harder and more time consuming. And that's when the security teams come knocking to pass their audit.
All it takes is an `npm config set` to switch registries anyways. The hard part is having a central party that is able to convince all the various security companies to collaborate rather than having dozens of different registries each from each company.
Rather than just a hard-coded delay, I think having policies on what checks must pass first makes sense with overrides for when CVEs show up.
(WIP)