Hacker News Reader

Comment by Someone1234 | Hacker News Reader

Let me provide context, since a bunch of people responding with "every package manager can be hit!!!" npm, by design, allows all packages to run package supplied arbitrary code as the logged-in user after an update completes.

That's an INSANE default. pnpm, by contrast, allows you to essentially "opt-in" only specific packages that need this (e.g. four out of thirty, in one of our projects). Then tacks on tons of other security settings, like minimum age, no trust downgrade, etc etc.

All attackers can attack packages by updating how a package functions; but npm is particularly problematic as it runs non-sandbox scripts as the calling user. Putting not just your project at risk, but your entire machine/network.

And this stuff has been known about for YEARS, they've taken no action.

View on HN

Furthering the idea that not all package managers are the same, there are entire cycles of the moon where I don't open nuget once. Some ecosystems simply don't need to vendor out very often, and these are the ones where you generally find the least news like this.

In about 99% of cases, I have the option to pick between Microsoft, a 3rd party or myself. I'm picking that first option every time I can. If M$ can't handle it, I'm hand rolling it.

Dapper remains the only constant 3rd party dependency in my projects. I don't know how much longer this will last with LLM assistance. The frontier models are very good at writing repositories over arbitrary sql schemas with low level primitives now.

> Furthering the idea that not all package managers are the same, there are entire cycles of the moon where I don't open nuget once. Some ecosystems simply don't need to vendor out very often, and these are the ones where you generally find the least news like this.

This however is only to some degree the package manager's fault. The JavaScript culture is strongly ordering tiny packages by individual people doing small things (left pad) rather than larger utilit libraries maintained by a larger community.

A larger community contributing to a larger library would mean that a larger community feels responsible and checks it.

That small package mentality a trace to web usage: JavaScirpt code is often sent to the client, not having a huge library but having small dedicated libraries means that it is a lot simpler for the bundler to not bundle dead code which is sent to the browser client.

With server side Node.js this lead to tons of dependencies ... which is worsened by npm allowing to have multiple versions of the same package in parallel. So if something depends on leftpad 1.0 and something else in leftpad 1.1 both are fetched and both are available.

This has been improving recently; one large project built on several heavy libraries that I've been supporting since 2018 currently installs ~180 dependencies without loss of functionality compared to how it worked, and what it depended on, back in 2018.

IIRC 6 years ago the full dependency tree congealed into more than 2000 packages. One small example is React itself:

- 5 deps: https://www.npmjs.com/package/react/v/15.6.2

- 0 deps: https://www.npmjs.com/package/react/v/19.2.6

Another is switching from create-react-app with its hundreds of transitive dependencies to vite, which, according to the test I've ran just now, currently has 15. Etc.

hmm maybe time to get into deno?

I mean, the current "allow ANY filesys operation" can't cope with modern supply-chain attacks...

with deno, you can specify folders/files that the execuble/library CAN touch (or CANNOT)

> That small package mentality a trace to web usage: JavaScript code is often sent to the client, not having a huge library but having small dedicated libraries means that it is a lot simpler for the bundler to not bundle dead code which is sent to the browser client.

Which is another part of this entire insanity:

Browsers are already <<huge>>. They're also built by <<huge>> companies companies that collect <<tons>> of analytics.

You'd think at this point they could present a proposal for a rock solid extended JavaScript standard library that would be based on actual website usage and would be comparable to what Java, .NET offer, obviously only keeping the parts that would be applicable to the web.

It sounds crazy but I think the Chrome installer is 150MB and an entire decent stdlib these days would probably be 1-5MB...

The web api is actually extensive. I can understand complaints about it being not exactly approachable, and some wanting a cleaner abstraction, but there’s no way that it is small. Most issues is about people wanting to download a small library than just vendoring the small snippet of code.

The other issue is the sheer amount of tooling and “plugins” for those toolings. Like the babel and webpack situation, which is truly kafkaesque.

That was circumvented once by CDNs hosting common libraries so that those would stay in browser cache, browser vendors then "broke" that by caching per origin. (So that an evil site can't detect whether a user had been on some target site before by testing whether assets are fetched from cache)

Issue probably is that the standards process is slow (unless it is a feature Google "needs") and full of bike shedding (which features and how exactly they'd look) and adaption of features by developers is slow.

JavaScript meanwhile should be stable enough as an environment to allow a broader standard library.

Luckily it is slowly getting better (see Temporal as new date library, replacing moment.js usage in many places)

How large a project do you typically use dotnet for?

IME dotnet dependency situation is a tire fire, not a month goes by without another dependency biting the dust or going fully commercial with no notice. Which is fair, I suppose, but Go and Java ecosystems don't have it nearly as bad.

I don't think going commercial has been that impactful. It sucks, it betrays the spirit of open source but whatever. A few examples:

- FluentAssertions had no moat, and it has been forked as AwesomeAssertions. Not sure what the author's play was here.

- Moq lost trust - we have NSubstitute

- AutoMapper and MediatR have been widely misused anyway

- Maybe MassTransit is a real bummer?

Switching to the forks/alternatives is still time and effort, often a lot of it.

> How large a project do you typically use dotnet for?

The largest dotnet project I am responsible for has around 50 megabytes of source files sitting on its main branch right now. If you include the generated WCF references it's probably closer to 100 megabytes.

> allows all packages to run package supplied arbitrary code as the logged-in user after an update completes

As opposed to the completely untrusted package supplied arbitrary code that the logged in user executes when they actually use the package immediately after installing it?

The package might not ever be executed on the user's machine. Depending on your setup, it might only be ran on a server, where the data that can be exfiltrated is completely different.

Sure but like.. come on. Is that really a defense? Most packages are run on devs machines. And it's not like "Oh it's just running on my production server, what could go wrong there" is any better.

We should not dismiss that it is slightly better. Production servers vary rarely have creds to the source repository nor to other production servers running possibly more sensitive code where investing in a smaller supply chain was justified.

Why you are downloading code if you're not even using it to run tests ?

And if you run tests in CI/CD, or in a container, why you are downloading code locally ? Only thing that comes to mind is code completion but surely most people at least run unit tests locally before pushing the code out ?

One malicious script that is run right after install vs one per each API entry point that might be called or not (transitive dependency).

You can't even install the package without running arbitrary code, that's quite different from most other package managers for languages.

Yes, this.

Regarding npm CLIENTS, PNPM is fundamentally different from (and superior to) npm or yarn.

Strongest possible recommendation to use pnpm.

It's also a good idea to use a private registry (eg via jfrog), acting as a proxy / pull-through cache, and point trad SAST and maybe AI scanners at it.

But dropping the npm client in favor of pnpm is a no-brainer. Speed, disk space, security, determinism, flexibility, fine-grained control over your dependency graph...

>Putting not just your project at risk, but your entire machine/network.

Between average hackers and extortion groups, foreign governments and state sponsored actors and last but not least my own government, I don't think there's much room left for non-compromised supply chains these days. Treat everything that can run foreign code as potentially compromized and keep everything compartmentalized. If you keep your crypto wallets or private banking info on the same machine where you do development, you're asking to get shafted one day. Or if you keep your big corporate github keys on the same machine where you do private weekend projects. It doesn't matter what you use in particular, even if some vectors are currently more popular than others.

If I can - I avoid NPM completely and just go with no-build projects.

One hour ago, while looking casually at a package.json, I saw this and was horrified:

  rm -rf pkg/snippets & rmdir pkg\\snippets /s /q & wasm-pack build --target bundler && node prepare-web.js

Looked like a strange mix of unix shell and msdos batch that would, on my box, try to rmdir "/s" and "/q". I asked Claude about this, and he replied something like "Yes that's a standard and clever hack to delete a directory that works both on linux and windows!".

Poor Claude has been trained on so much awful human code that it required several prompts for it to admit that there was indeed a problem.

The industry is the process by which convenient crap like this gets standardized.

Yikes. I would never approve a PR with that in it.

Your agent might!

> Poor Claude has been trained on so much awful human code that it required several prompts for it to admit that there was indeed a problem.

Claude probably birthed this abomination in the first place

I hope the "standard" part is as much of a hallucination as "clever" is

To meekly defend the indefensible here: it's not like rmdir on Linux (I won't speak for all Unixen) can cause loss of data, since it only removes empty directories.

In this case, the rm -rf before that does. The rmdir is the Windows command in this example and with /s /q, it will quietly delete everything.

Yep, but that could still cause issues (those entries could be used as signals, or be mount points for currently unmounted partitions, etc). rmdir anything that start with "/" should be an absolute no-go.

To say nothing about running a sequence of shell commands without the -e option.

> since a bunch of people responding with "every package manager can be hit!!!" npm, by design, allows all packages to run package supplied arbitrary code as the logged-in user after an update completes.

This is semi-common and in no way unique to NPM.

And even in the ones that don't, having to wait until the project executes to begin its attack is a minor inconvenience for malware.

What other package managers do this? I don’t think Ruby does

https://docs.ruby-lang.org/en//master/Gem.html#method-c-post...

Python does too I believe.

Really the reason not to allow that is for robustness, not security. You ideally don't want package installs doing random stuff to your system because package authors are generally bad at doing that sort of thing cleanly.

The security impact is relatively minimal because as other people have said, you just installed a package. What's the very next thing you're going to do? Compile/run it obviously.

A lot of packages are pulled in to call minimal bits of the actual library. I obviously don't have any statistics on this but my instinct would say that for the average application only 5% of an average package is actually used.

So not running package installation scripts is a huge, massive problem.

So what? Packages can just put their backdoors in some initialisation code that is always used.

It is possible that not running package installation scripts could improve security, but for that you need really good sandboxing/compartmentalisation of library code, e.g. with CHERI, WASI component model, or if all of your code must run in a secure context it probably helps.

But those situations are unfortunately rare in my experience.

It doesn't matter how much of the package you use. Here, you can use literally 0% of Koa and get pwned by one of its transitive dependencies (koa > cookies > keygrip > tsscmp) by simply importing the parent package:

    mkdir demo && cd demo
    npm install --save koa@3.2.0
    echo 'console.log("--- pwned by a transitive dependency ---")' >> node_modules/tsscmp/lib/index.js
    node -e "import 'koa'"

--- pwned by a transitive dependency ---

My point was for proper package management tools that don't allow running scripts.

His example did not involve running any post-install scripts.

Most of them? Ruby gems have hooks, Python has setup.py, deb, rpm have them too (relevant if you're installing from 3rd party sources). Elixir/Mix doesn't technically execute code on install, but your language server builds the dependencies as soon as you open the project, which can execute arbitrary code.

Either way it misses the point, nobody just fetches code and removing post-install scripts wouldn't change much because you're going to run `npm run something` 5 seconds after you run `npm install`.

You're right. I said the same thing and got downvoted too. Don't let it discourage you.

> That's an INSANE default.

I agree that not running arbitrary installation scripts is the right default, but it's just an incremental improvement.

The practical difference between code that runs at installation and code that runs when the package is executed is, very typically, a small amount of time.

IMO, the hyperbole here hurts because it distracts from more effective efforts.

> IMO, the hyperbole here hurts because it distracts from more effective efforts.

For example?

Can I just run npm update diff and see all changes across all updates compared to the last reviewed code in node_modules? Why not?

> they've taken no action.

Not running lifecycle scripts by default is eventually going to be the default behavior. Late is worse (edit: I meant better) than not at all. https://github.com/npm/rfcs/pull/868

Wait how is being late worse than not doing it at all? Is it true for mortgage payments and apologies?

Mosts packages manager, allow that.

pnpm can still be exposed, afterall the worm simply have to wait you run tests locally.

I suppose.

But that's a "Perfect is the enemy of good"-like argument. Wherein: Why even reduce an easy to exploit attack surface when there could be holes elsewhere?! Because, you know, it makes things much more secure even if imperfect.

Plus, to me, it is a culture issue. npm just doesn't take security seriously, so we don't see these improvements, and if there was additional test hardening later, I don't expect we'd see them in npm either. Since, they just don't care.

The biggest problem is not software but culture, not at npm, but in the js ecosystem. The js ecosystem is simply a juicy targets, the attack surface is enormous. The attacker can make their attack more sophisticated, there will always be a maintainer that can seed the worm spread.

Meanwhile in the nuget ecosystem is way smaller and have way less mainteners involved for a single given dependency.

I'd go further and say that how JS and the web itself has been run over the years has predisposed it to this sort of thing.

JS didn't have a passable stdlib until ES6. It had bugs built into it because Eich was given a stupidly short time window to deliver the first version. Everyone (particularly MS) had (and still sort of do) their own way of interpreting the language. In spite of all of this it became the primary way of developing applications for public consumption.

This led to a bunch of people who wanted to be the 10x JS engineer to solve problems with their own libraries and technologies. None of them really talked, they just threw their packages on NPM's registry without second thought and some gained widespread use just by accident.

Google tried fixing some of this with Dart but chickened out at the last second. TypeScript was designed by someone competent but can't fix the larger cultural issues.

This is what happens when you put SV hubris and "moving fast and breaking things" over doing things the right way.

Yes this is kinda my point. Instead of having a few projects/org, it's a constellation of packages too small, it's impossible to know who you depend on when adding a dependency.

> Why even reduce an easy to exploit attack surface when there could be holes elsewhere?! Because, you know, it makes things much more secure even if imperfect.

I'm still trying to calibrate my take on this view.

If attacks are randomly chosen from the set of all potential vulnerabilities, without the attacker knowing which ones had been patched, then that logic clearly makes sense.

But in an adversarial situation where the attacker can guess which vulnerabilities you still have unpatched, or can try many different attack vectors, then having already patched some other vulnerabilities doesn't matter so much.

I guess reality is more complicated though.

You can isolate it through bubblewrap; I moaned about it here and there's no point in repeating it:

https://news.ycombinator.com/item?id=45041798

If you only ever use js/ts for frontend projects (like we do), it closes one major hole that I'm aware of, which still leaves at least two:

- the editor possibly starting random binaries from inside the mode_modules (such as biome, vitest, tsgo)

- escape from sandbox by using some kernel vulnerability, of which there have been many recently

They have taken action as of very recently. The latest version [1] of npm warns when there are install scripts and tells you they will be disabled by default in a future version, with a per-dependency opt in mechanism [2].

[1] https://github.com/npm/cli/releases/tag/v11.16.0

[2] https://github.com/npm/rfcs/pull/868

This is way too little, way too late.

To see what I mean, try actually packaging a cross-platform binary dependency in their ecosystem.

I have; you specify one optional dependency per platform and set the requirements in each package. It works fine. A bunch of packages do this (e.g. esbuild). I don't know what your complaint is or what you're asking for.

npm is so bad at everything it's insane.

SIXTEEN YEARS of development and they can't even resolve a tree of dependencies in the correct manner unless you nuke the lockfile and node_modules.

Dependency resolution is literally the number one task and they fail at it. How can you expect them to be good at anything else? Absolute joke.

Maybe NPM is scared to break a ton of packages? I also think action from NPM on the repo level is vital

I went through the package.json on my machine - seems like ~400 / 60000 or 0.7% have (pre|post)install. (That's not all of the scripts that run at install)

Seems to me like a backwards compatibility is a non argument since pnpm is popular enough to stand as existence proof that scripts can be, at least, opt-in

IMO - pre- and post- install scripts should just be abolished/deprecated. It should require a special dispensation from npm to even publish one. A better system for binaries (needed by esbuild) is probably needed.

Even saying "just use pnpm" isn't enough, we need to get the developer community to herd immunity and that isn't going to happen on an opt-in basis.

I would love for npm to sandbox as well. But I think the better way forward is just turn off scripts.

So do the pre- and post-install scripts of Debian packages. The problem is not this but the lack of verified and controlled release channel.

I think another thing that affects security is that in javascript culture people often tie to the latest version instead of concrete version.

This makes it so an update to a popular library can compromise a huge number of packages that depend on it.

In Java for example almost all packages specify a concrete version, even if someone compromises the latest the blast radius is usually pretty small.

MS Nuget is also lock-by-default. Latest-by-default should be considered harmful unless the package manager is directly vouching for the veracity and reputability of the packages.

NuGet is lock-by-default for the parent package, but with the move from packages.config to <PackageReference> it's no longer lock-by-default for dependencies.

It never made sense the other way. If I reference a package, logically I'm also referencing its dependencies at the version that the package uses. Forcing the user to also reference dependencies of dependencies of dependencies means the package reference lists aren't DRY.

But just the dependency list isn't sufficient to pick a specific version, thanks to dependency ranges. If Package A depends on Package B >= 1.0, and Package B has v1.0 and v1.1 available, it will use v1.0. But if Package B suddenly unlists v1.0, then future restores will change to v1.1.

Ah, I see the worry. A supply-chain attacker can use de-listing to force an upgrade to the malicious version if clients have dependency ranges that reach into the future.

I didn't know about that one.

In general, any dependency system that allows "you can silently upgrade to versions of the package that did not exist at the time the packagereference list was created" seems to be a vulnerability.

It's frustrating since this vuln seems trivially simple to fix, at a glance... although it would require an API change in PackageReference. Mandatory lockfiles by default, or getting rid of the floating versions misfeature. BindingRedirects let you override declared dependency versions anyways, they're not a blood pact.

It seems trivially simple until you have two dependencies with conflicting exact version requirements... So I don't think you can get rid of floating versions entirely. They did add NPM-style lockfiles for PackageReference, but currently not mandatory.

The version numbers for BindingRedirects are orthogonal to the package versions. You can have multiple package versions use the same AssemblyVersion so that applications don't need to create BindingRedirects. (e.g. Newtonsoft.Json - 13.0.0, and 13.0.1 in NuGet are both 13.0.0.0 for binding redirect purposes) And .NET Core/5+ don't need BindingRedirects at all!

Won't pinning a version lead to dependency hell, not to mention potentially using vulnerable versions if you don't a new version after it has some CVE fixes ?

> That's an INSANE default.

It's also the standard, and by far it's the contrast to not allow this. pnpm has a massive advantage of being the non-standard package manager, npm does not have that - what do you suggest that npm does?

There are so, so many things that NPM could do.

It could require a 48 hour cooldown period on any package update that wants to add an install script that didn't have one before, and has a certain number of downloads. And it could publish the list of these so security researchers have an opportunity to scan them.

It could add an optional key to package.json that allows someone to whitelist which packages can run install scripts.

It could add a Hardened Security program where (1) package maintainers could opt into a program where multi-factor confirmation by maintainers is required on every publish, even those triggered by CI; (2) this hardened package status would be public, and (3) a developer could set a flag in their package.json that causes any npm action to act as if all non-hardened packages had frozen versions.

And so much more.

You realize that "dependency cooldowns" as a popular concept are extremely new, right? npm manages the installation of dependencies for millions upon millions of users across the globe.

> It could add a Hardened Security program where (1) package maintainers could opt into a program where multi-factor confirmation by maintainers is required on every publish, even those triggered by CI;

Great, they did this.

> And so much more.

This shit takes time. Yes, they should have done this on day 1. Acting like any of this is easy to retrofit is just nuts though.

What is being said is that a new flag like '--minimum-release-age' would take, realistically speaking, tops 4 hours to implement (without AI assistance), plus a good 1 week of thorough testing, and maybe a 1 month period of progressive deployment. Come on, let's give it a total of 1.5 months, for good measure.

Of course this should have been started since the beginning of the major recent stream of supply chain attacks, circa 2024 or 2025... but even assuming the most backwards calendaring possible -starting after the last bug compromise (Axios, on March 31st)- that new flag should have already been shipped a couple weeks ago.

Shit does take time, but where there's a will there's a way, and nobody buys that this shit would take that much time.

Have you ever managed software as critical and ubiquitous as npm?

Not infra, but final product. I know, corporations move slow. But when there is a critical issue, and an actual desire to solve it from someone in a suit, suddenly turns out that the cogs were always able to speed up and move fast...

i've been thinking about this as well. but having built a startup, i've learned that users don't care as long as they are given the value and most convenience. they don't really care much at security as much as we do. just look at openclaw? but maybe it's our job to make sure it is taken care of vs assuming the user cares and just make it look seamless.

users don’t care about security until it goes wrong. then they will be angry.

security is a hidden requirement.

> Let me provide context, since a bunch of people responding with "every package manager can be hit!!!" npm, by design, allows all packages to run package supplied arbitrary code as the logged-in user after an update completes.

Many package formats before NPM allowed for it, and frankly, it matters little, because if it can add code to your app it can run malicious code. The fact it executes on package install rather than when dev runs tests or the app matters little, and in general if environment is sandboxes, the package install is also ran in the same sandbox so disallowing it changes little.

so yes, every package manager can be hit, the reason is twofold

* JS is such a lowest common denominator it has that much more clueless users so just by scale every issue will be more common than in other languages

* extreme fragmentation leading to hundreds of packages needed for even small projects, which is again more chances for compromise

Every package manager, by design, allows arbitrary code execution after the update completes. It is the entire purpose of a package manager. There is no point installing code that does not run.

So does dpkg & rpm

That usually has a separate maintainer checking things, not updating automatically and often being the author of those packaging scripts, as they are often distro specific.

Nearly every package manager I've ever used had post-install scripts. Most run as root, since that's what usually what the package manager runs as.

It's not unreasonable: you're already installing software, which presents risks. If post-install scripts were not a thing, a payload could still run because you ran the software you installed. Or because the installer added it to auto-run. Or because the installer placed it somewhere where it would be dynamically loaded all the time.

Nearly every package manager I've ever used had post-install scripts.

You're collapsing two different threat models. The risk isn't that code runs, it's WHEN it runs. This worm spreads because npm install runs arbitrary scripts as you, automatically, just from resolving the tree. You don't have to build it, run it, or even import it. Opening the project in an IDE is enough. apt/dnf scripts run on packages a maintainer signed and a distro gatekept. Not on whatever some rando pushed to a public scope 20 minutes ago that landed in your lockfile six levels deep. "They both technically execute code" is true and beside the point. One runs signed code from a trusted path, the other runs unsigned code from the default automated path. That's the whole ballgame.

> You're collapsing two different threat models. The risk isn't that code runs, it's WHEN it runs.

> You don't have to build it, run it, or even import it

If you just installed something with npm, chances are you'll be running it shortly, either as a tool or a library, probably minutes or seconds later. I imagine the use case of installing an npm package you don't plan on using or transitively importing, constitute a small portion of npm installs.

> apt/dnf scripts run on packages a maintainer signed and a distro gatekept

Unfortunately apt/dnf isn't much better here because random tutorials online suggest people add random repositories where the creator of any repository effectively has root access to anyone machine that adds it as a remote.

Don't add random repositories from random tutorials? Come on, it's basic Internet hygiene. Entirely different thing.

It's the exact same problem when random tutorials (and official pages) recommend to do a curl "URL" | bash to install something. Every time that I see it, I look it suspicious.

Most package managers with postinstall scripts are also heavily curated and have reputation systems. As you say, they run as root, so the high trust requirement is definitely warranted. Anyone can upload an npm package.

I think it’s just a bundle of issues. Deep graph of dependencies, distribution of minimized code (java has jars, but I don’t remember scripts), and nearly impossible to audit. With most projects in other ecosytems, you only have to interact with a few developer/orgs. But with npm, you add one library and you need to essentially trust 10s of entities on the internet.

That's why we don't let the developers run system package manager install scripts as root. We do let them run npm inside containers, which is still more access than I'd like them to have.

Is this one of those places where it takes 6 to 12 weeks to get something new installed on your machine?

You mean directly on the machine? Not in a container? That would be a recklessly fast timeline. The configuration control board meets quarterly and it usually takes 4 or 5 meetings to clear a piece of software.

Interesting. How about in a container? What industry is this, just curious?

Defense. Containers on our normal network are the wild west, but anything that touches the secure network has to be signed off on by the clients.

Reply on news.ycombinator.com