AURpocalypse now: a look at the recent AUR attacks

https://lwn.net/SubscriberLink/1077619/f7b07c5489fdd43a/

by jwilk • 1 month ago

134 101 1 month ago

lwn.net

rvz• 1 month ago

Who still uses Arch btw after this?

akerl_10• 1 month ago

Is there another distro that has an equivalent of the AUR with handling you think is preferable?

orbital-decay2• 1 month ago

AUR is fast and loose and doesn't do much "handling" by design, so it's hard to find any equivalent repo. But there's always a tradeoff between fresh (nixpkgs unstable, might be the closest) and tested (Debian).

guilhas• 1 month ago

Gentoo

But let's hope we get this solved, like peer review model, vouch, or something

It is very good to be able to find build/install files for everything

dokyun• 1 month ago

SlackBuilds.org is pretty sensible.

dizhn• 1 month ago

Opensuse OBS. Tiny bit better because the build environment doesn't allow a network and binaries are not allowed as far as I know. Fedora has a similar thing COPR. Both of these support building packages for other distros as well as appimage, flatpak etc.

With opensuse official packages also use the same infrastructure. It is actually quite fascinating and powerful. (I know a lot less about COPR but I would imagine it would be equally as good. Wezterm switched to that for its packages)

einsteinx22• 1 month ago

No because there’s no way to handle an open submission repository at all. It’s impossible by design since anyone can submit packages to it.

I would never use anything equivalent to AUR on any distro due to the obvious security implications. That’s been my position for as long as I have known about Arch. I never understood Arch users using the AUR as a selling point for the distro.

Then again I live in the opposite end of the spectrum where I run only Debian Stable on my Linux desktop as well as my servers, where packages make it through Sid and Testing before getting to Stable and I can be relatively sure any supply chain attacks have been caught by then (like xz for example which was caught before it left Sid).

For those unfamiliar with Debian, Sid is basically a rolling release similar to using Arch with the official repositories (which is already dangerous without even touching the AUR), then packages move to Testing, then later eventually make it to Stable.

tormeh2• 1 month ago

Windows where you just download .exe and .msi files from random websites. Not a Linux distro, but the closest security model I can think of.

giancarlostoro• 1 month ago

I still do, I just don't touch AURs anymore.

anagram666• 1 month ago

If you want something from the AUR, just don't be lazy, read the pkgbuild.

segfalt_• 1 month ago

I do, I'm just choosy about aur packages I use

rcxdude• 1 month ago

The AUR has consistently had warnings around it of 'verify the PKGBUILD', far more so than any other package repository that allows anyone to sign up. Probably the only notable difference is the ease of taking over an orphaned package.

zbentley2• 1 month ago

The AUR is not the Arch package manager or repository. The main Arch package repos are managed similarly to Debian, or Fedora, or whatever--caveat Arch's nature as a rolling release, but in terms of vetting and ownership/security, the approaches are similar. pacman installs from regular, real, vetted repositories by default. pacman will never install from the AUR. pacman is the official Arch package manager and the only one that is provided with the main Arch distribution/install instructions.

The AUR is, as many others have pointed out, a deliberately un-vetted pile of random Git repos. Arch deliberately doesn't even ship with a default one-click installer for AUR packages; their published guidance is "git clone this stuff from wherever it's hosted and build it at your own risk". Plenty of third-party, non-Arch-blessed tools turn that into a one-click process, but it's not "part" of Arch itself--at least not any more than, like, curl | bash or directions on how to add rando websites to /etc/apt/sources.list.d is part of Debian and friends.

I've used Arch as a daily driver for years. At peak, I've had five (5) total packages, with no transitives, installed from the AUR. Today I have one: sublime-text-4. It's perfectly possible--and extremely reasonable for many users, even power users--to live in an AUR-less world, or to use so few AUR packages that the guidance of "read what you're installing, doofus" is manageable and not onerous.

QuaternionsBhop• 1 month ago

I was not affected

beej71• 1 month ago

I do. I just keep reading the diffs on the PLGBUILDs.

AshamedCaptain• 1 month ago

I'll note that OpenSuse also has Packman which a shitton of people enable (for codecs), has also 'one namespace only' an looser policies than the main distro.

I do not think this something you can escape by switching distro.

cqz• 1 month ago

Yes, the only reason this isn't happening in other distros is simply popularity.

Namespacing is the solution, and as mentioned in the article some ditros do indeed have namespaced user repos, like Fedora's Copr. The trust model of a flat namespace user repo is completely broken when the maintaining user can change at any moment.

PalmPilotProMax• 1 month ago

Isn't Arch's AUR flat namespace quite unique? Ubuntu's PPAs are also not flat.

badreligion42• 1 month ago

Packman is more akin to rpmfusion, than AUR. OBS is the AUR equivalent for OpenSUSE.

isityettime5• 1 month ago

"One namespace" is also technically true but doesn't work the same way with dnf or zypper as it does with pacman. dnf and zypper both make it easy to be explicit about the priorities of your repos and also to track which packages come from which repos and prevent that from changing. Plus openSUSE has a generously free public instance of the Open Build Service that you can easily use to host your own repos, and which hosts many individual repos you can add for specific purposes. When I ran openSUSE I always just ran my own repo there with only the extra packages I actually wanted, often just "forking" packages from repos hosted by well-known openSUSE developers so that I didn't have to manage updating the source packages myself but still didn't pull in the whole world from those repos and also didn't implicitly trust anything as loose as the AUR.

OBS is more like Ubuntu's Launchpad or Fedora's CO0R than the AUR. Random strangers can't take over the packages of others just because they go idle, and it's a bunch of separate repos, not one. Totally different trust model.

jjmarr• 1 month ago

I use Gentoo. You have to specifically install "overlays" and every package maintainer would make their own overlay. You can't easily take over an overlay without the original person's permission.

That being said, still one namespace. Once you add an overlay it can replace any package it wants.

It's also Gentoo so too hard for most people to figure out.

isityettime5• 1 month ago

Zypper at least has a notion of "vendor", so you can arrange things so that only the handful of packages you care about will actually come from Packman.

Ubuntu actually has first-party repositories with proprietary codecs.

Nixpkgs is a pretty comprehensive monorepo of packages with a more normal review process than the AUR, and it includes non-free software as well, plus the model with flakes for third-party stuff is that you trust individual publishers for their little repos rather than one giant grab bag repo of unreviewed content like the AUR.

RPMFusion for Fedora kinda has a similar profile, in that it's a shared repo for various things unsuitable for the main one, but it follows more or less normal Fedora packaging and review standards, doesn't it?

Supply chain attacks are possible everywhere and some distros have particular weaknesses, but the AUR really is pretty much uniquely bad here.

eptcyka• 1 month ago

Nix also forces builds to be sandboxed. Now you actually need to run an infected build output to be affected.

nickjj• 1 month ago

In case anyone missed it, the latest version of yay (v13+) supports being able to skip recently added packages through its new Lua extension system https://jguer.github.io/yay/lua.html#upgrade-selection-hooks. You can control the threshold since it's just user configuration now.

A bunch of common yay commands also return back the last updated time of a package thanks to https://github.com/Jguer/yay/pull/2846.

dSebastien• 1 month ago

Hope Omarchy will make that a default:)

kdeldycke• 1 month ago

The thing is that hook is not enough: `UpgradeSelect` only applies to `yay -Syu` so it only filters the upgrade list.

Nothing protect you from a `yay -S foo` install and its dependencies. So this is not a guarantee or enforcement of a minimum release-age.

Actually writing this reply I went ahead and pointed that out in an issue at: https://github.com/Jguer/yay/issues/2883

vlovich1234• 1 month ago

The issue you linked isn’t anything to do with the issue you’re describing. Wrong one referenced?

AquaWeasel• 1 month ago

Despite that official Arch repos weren't affected in this attack, I would not recommend using Arch (or any rolling release distro) for anything that requires security. (Imagine if the xz backdoor targeted Arch...)

An Arch maintainer that I personally know once admitted that he rarely review upstream changes when bumping package versions. He only does that when the build breaks.

I can't blame him for what he did, since it's not reasonable to ask package maintainers to spend all their time on those stuff, especially in this "Age of AI" where more and more software are being aggressively refactored (or rather rewritten) and added more features.

What we can do is choosing a stable distro (like Debian) where packages are more thoroughly reviewed, and apply security practices (such as TOTP, sandboxing browsers and video players, etc.) even though they cause inconvenience.

zbentley2• 1 month ago

An Arch maintainer that I personally know once admitted that he rarely review upstream changes when bumping package versions.

Cool story bro. Assuming that's common, I have trouble understanding why Arch (non-AUR) is any more at risk than Debian--besides the latter being more popular and having more users/incidental testers, which is a real benefit if that's your goal, but has its own drawbacks (like older and known-vulnerable packages lingering for longer before updated releases are made available).

it's not reasonable to ask package maintainers to spend all their time on those stuff, especially in this "Age of AI"

Aren't Debian and friends similarly at risk of this as well, then?

security practices (such as TOTP, sandboxing browsers and video players, etc.)

I'm not sure if those are more or less prevalent on Arch; I know that many IDEs and GUI programs I've installed on Arch ran by default in Flatpaks or similar, and Debian/Ubuntu like Snaps, but I'm honestly not familiar with whether those ecosystems have significant and/or equivalent penetration in different distros.

skjfjnflw• 1 month ago

Debian freezes the package versions on release of each Debian version and then cherry picks critical fixes for the rest of the Debian version's lifecycle. So even if they never review the code (and I don't expect them to), they're less likely to release malware before it's discovered by others.

Ferret74462• 1 month ago

I don't think Arch maintainers are responsible for auditing upstream. They package the upstream only.

ChocolateGod2• 1 month ago

If you package software for a distro, you have some responsibility for reviewing what you publish.

If you distribute an update that has malware, that is you publishing malware.

akerl_10• 1 month ago

where packages are more thoroughly reviewed

Where are you drawing your conviction that non-rolling-release distro maintainers are doing a more thorough review of upstream changes?

such as TOTP

What?

matheusmoreira• 1 month ago

Alarming. The whole point of official repositories is they're trusted, since they are maintained by trusted people. If one can't trust maintainers to be responsible, then the official repositories are no different than the AUR.

kryptiskt• 1 month ago

It's a complete fantasy that Debian maintainers do a thorough review of changed packages. It's not a responsibility they have, and it would be impossible anyway (how many packages are there in Debian? How many maintainers are there?).

Ferret74462• 1 month ago

Devil's advocate, except partially serious.

This is a good thing, because the warning about checking everything you download from the AUR, which has always existed, is now actually "enforced". People respond to consequences.

cookiengineer• 1 month ago

Note that the AUR attacks were part of the larger miasma worm campaign, gradually trying to gain more control through various package ecosystems since the RedHat prototype campaign.

Mitigation Tool: https://github.com/cookiengineer/antimiasma

Blog Post with details: https://cookie.engineer/weblog/articles/malware-insights-mia...

cozzyd• 1 month ago

I love the smell of npm install malware in the morning.

stephan-cr• 1 month ago

Yes, feels like in those cases npm and bun are not far away. Coincidence?

scott_w• 1 month ago

Simplicity of the stack I think. I don’t think this is an npm-specific issue as the attacker could also download a bash script and run that instead.

MintPaw• 1 month ago

A side note, isn't package maintenance something that can actually be solved to some extent by LLMs? The prompt would be something like "Clone this repo and build this package while building/bundling as few other packages as possible with minimal code changes."

Then set it in a loop on all the packages for a particular system, I don't have experience in package maintenance and would be curious what kind of issues would come up.

isityettime5• 1 month ago

Packaging for Linux distros is about review, following standards and conventions, authority, and responsibility. (And maybe also sometimes patching for compatibility.) LLMs can sometimes help with some of the mechanical parts, but the curation and trust stuff not so much.

OJFord• 1 month ago

And everyone does this individually you mean, rather than sharing the result as 'a package'?

It's an interesting idea, not sure I'd recommend it broadly, but if someone told me they prefer to trust an LLM than third-parties I'd get it at least.

jchw3• 1 month ago

The AUR really has been known to be low-hanging fruit for bad actors, which makes it somewhat surprising it took this long for it to be taken advantage of.

I have many opinions regarding this situation, but it mostly doesn't matter. AUR staff and AUR helper developers will figure out what they want to do, hopefully they will find a good approach.

But what I personally take away from this is simply that it has become worth it to target desktop Linux with malware. Or at least, moreso than previously. It is perhaps a good sign in some ways that the desktop is starting to be taken more seriously.

The bad news, of course, is that the Linux desktop is a bit of a train wreck in terms of security hygeine. It's getting better, and Linux does have the advantage of having some powerful primitives to exploit, but the desktop suites come from a totally different world, and I fully expect we'll also see more malware propagated through KDE's New Stuff integration (which goes through Pling.)

Normal_gaussian• 1 month ago

I'm not sure if it is that the desktop is being taken more seriously, or that its easier to write code that works on many distributions and configurations, greatly reducing the cost and increasing the value of the existing 'market'.

npodbielski• 1 month ago

I would say that it is now very easy to steal 'AI' providers credentials this way. And then you can use them to write more malware or scam or use models for generating speech to call people and get them to 'redeem'. Or at leat to me this seems more sensible than injecting just malware.

exceptione• 1 month ago

  > It's getting better, and Linux does have the advantage of having some powerful primitives to exploit, but the desktop suites come from a totally different world,

When opening the printer configuration page in the KDE configuration panel, I was pleasantly surprised to see it's process runs wrapped inside a bwrap session. Cups is a bit of old and dangerous; I'm glad they sealed that off inside a sandbox. If you ask me, I would make this approach the standard for any software. The configuration panel for fonts doesn't need network access, so at least `bwrap --unshare-net`

graemep• 1 month ago

I think KDE's approach is a greater danger. They both come with warnings, but with AUR (depending on tools) allows you to inspect the PKGBUILD. KDE just gives you a warning and no easy way of looking at what you are installing, it is not clear what contains executable code, and its enabled by default.

In general things that are not part of your distro's supported repos (KDE's AUR, language package installers like npm and pypi, Ubuntu PPAs, etc.) seem to present far more of a risk.

vlovich1234• 1 month ago

But what I personally take away from this is simply that it has become worth it to target desktop Linux with malware. Or at least, moreso than previously. It is perhaps a good sign in some ways that the desktop is starting to be taken more seriously.

Absolutely the wrong conclusion:

1. This is a super low hanging fruit attack

2. The ROI is significantly higher than the cost of attack

3. The cost of the attack is now drastically reduced due to LLMs (not that they necessarily mattered here but hard to rule out).

It’s less about the Linux desktop where Ubuntu is dominant and more that Arch security is so bad regarding how they maintain the AUR. Seriously it’s worse than Swiss cheese in terms of putting up roadblocks.

embedding-shape6• 1 month ago

that Arch security is so bad regarding how they maintain the AUR

I still don't understand what people expect Arch to do here? It's a user-contributed repository open for anyone, Arch maintains their own official repositories that are separate from AUR, what would need to change in the maintenance of the AUR for you to consider it to be "properly run" or whatever, and it doesn't stop serving its core purpose anymore: to allow any user to upload any PKGBUILD?

If you're advocating for not allowing users to upload their own PKGBUILD anymore, then it stops being the AUR, and for the people who agree with that, they can already exclusively use the official repositories instead of touching the AUR.

In true Arch fashion, the security is up to you here, review stuff before installing 100% unknown and random 3rd party software from the internet, as always. If you don't want to review anything? Stick with official stuff, and you won't have issues.

madduci• 1 month ago

I don't know how long will it take to attack also flatpak/flathub and snap stores as well.

jchw3• 1 month ago

Snap has had a couple of attacks, at least. I recall one that was a fake crypto wallet. I'm sure Flathub has, too.

That said: the attacks on Snap and Flatpak are bound to be less interesting; it doesn't share the same mechanics that make AUR scary. I still hope that Linux infrastructure vendors work to come up with good ways to "raise the floor" in these cases.

orf2• 1 month ago

New user registration was stopped on June 11 and then re-enabled after the project added Anubis to try to foil the attacker's mass account registrations. That did not work

This confuses me - why would a proof-of-work anti-scraping system like Anubis prevent registrations?

lemagedurage• 1 month ago

Implementing a bot to do registration is not a "copy as curl " click away anymore, and creating millions of accounts maybe become computationally expensive.

This is not much, but it could deter some low effort adversaries.

orf2• 1 month ago

I see that, but the adversary wasn’t a low effort one, and didn’t need millions of accounts.

stevefan1999• 1 month ago

Well all of those attacks are just supply chain attacks, and it is basically exploiting people's trust. With LLMs, the speed and velocity of pumping out malice raised are now significantly faster.

It is so sad that every goodwill eventually got enshittified as well.

BoingBoomTschak2• 1 month ago

Fantastic and anti-sensationalism write-up from LWN, as usual. They continue to deserve my monies.

ChocolateGod2• 1 month ago

The AUR is effectively a pastebin for PKGBUILD files.

Some people (many of which don't even use Arch itself) have been treating and advertising it as something different. It's not a software distribution method meant for normal computer users.

cbarrick• 1 month ago

And for those that aren't aware, a PKGBUILD file is just a bash script.

maxmax33• 1 month ago

Humble question: how do you find out if your system has been affected by a malware?

I know that for AUR there was a specific list of affected packages (that I checked, and haven't installed any of them), but I'm interested more in a general way. It could be from AUR, npm, or many other sources. Some malware could break and lock immediately the system, but other could stay there silent for months, so how to find out if there is any?

I haven't run an antivirus since I last used Windows 20 years ago.

regexorcist• 1 month ago

You might start with ClamAV and something like Little Snitch.

samtheprogram• 1 month ago

Would love to know the rebuttal / why you were downvoted.

embedding-shape6• 1 month ago

In theory, you'd mostly care about exfiltration of data, so watching/actively managing exactly what network connections your computer/network can do, would give you upfront notification when it happens, but of course not earlier. And if your root/hardware somehow is infected, your monitoring/management tools might be affected too, so then you're basically out of luck except with external network gear outside of your computer. But then those could be infected too, and so on.

Ultimately I'd say it depends on your risk profile, but using something to actively approve/deny network connections on your local machine, is a great start that'd defeat most of these simpler "exfiltrate information ASAP" malware that seems popular at the moment.

diogenes_atx3• 1 month ago

you'd mostly care about exfiltration of data, so watching/actively managing exactly what network connections your computer/network can do, would give you upfront notification when it happens

If you have a list of good CLI utilities, you could run them in a bash script (e.g., network-monitor.sh), which would run in the background, and then redirect the output data to another file (e.g., network-monitor.txt). The key concept here is "baseline" -- you need to know what normal baseline network activity looks like, so that you can identify anomalous behavior. The way to establish a baseline is to gather a lot of data from the system.

The following are a few useful command line utilities to use for a host intrusion detection system (HIDS) using a simple network monitoring bash script. However, I am not sure exactly how to tweak the options. Also need to find a way to check for data exfiltration:

-list open ports and processes that own them: netstat -lnp

-show open network ports: lsof -i; netstat -an | grep -i listen; netstat -nap

Then it would be relatively easy to write a python script with regex tools to parse the network-monitor.txt file, establish a baseline, analyze the data for patterns and search for anomalous behavior.

Besides network monitoring, there are other command line utilities you can use to check the system for possible intrusion, which you could run in a separate bash script as part of your Host Intrusion Detection System (e.g., hids-users.sh):

-show members of root group: cat /etc/group | grep root

-show users logged in: w #if you are the only user, you should not see more than one account logged in

-search for all accounts with UID of 0: grep :0: /etc/passwd #ideally there should be only one UID of 0 on the system, but attacker can create more.

-check that daemons who never log in have * or !, meaning no passwd: cat /etc/shadow

-look for orphaned files, possible sign of attacker temp account deleted: find / -nouser -print

-search for new user accounts that are not part of regular build: sort -nk3 -t: /etc/passwd #sort numerically third column (UID), colon delim (-t:)

EDITS: several small changes for clarity and also in response to comment below

TacticalCoder• 1 month ago

Some malware could break and lock immediately the system, but other could stay there silent for months, so how to find out if there is any? > I haven't run an antivirus since I last used Windows 20 years ago.

That'd be the role of an IDS (Intrusion Detection System). Things like file signatures of your entire system being saved to another machine (for example an offline/airgapped one) beforehand (for example by plugging your main machine's SSD as a secondary drive on the airgapped one).

If you suspect shenanigans, you take your machine offline, you remove its SSD (your BIOS/UEFI is also a concern), plug it to your airgapped machine with the IDS: it compares all the files (binaries, config files, etc.)' checksums with the past ones.

It's a bit of a lost art but it could make a comeback seen what we're facing, now nearly on a weekly basis.

Some distros have a way to check for file integrity as part of the package manager: but you can't trust the infos coming from the machine itself if it's been compromised.

mentalgear• 1 month ago

The funny thing is: with FOSS you know what's going on attack-wise as they are transparent , but there must be huge upheaval in the closed source (e.g. MS) software that is just kept quiet by stock-price-chasing executives.

ameliaquining2• 1 month ago

A lot of people in this thread seem to be treating this situation as a referendum on the security of package repositories that allow anyone to create a package. Possibly because that's interesting to more people, since npm and PyPI are more widely used than Arch.

But unless I've badly misunderstood something, the key thing that made this attack possible is this "orphaned" thing that lets you grant write access to an existing package to the first person who claims it, without any control over who that is. I don't see how this could ever be a safe thing to do, I'm not aware of any other package repository that has it, and I struggle to guess what whoever built it was thinking. If AUR just turned off that misfeature, they wouldn't be having this problem.

(The article quotes someone involved as saying that the "orphaned" feature is good because good actors can also use it, but that seems irrelevant if it also opens up an unmitigable machine-takeover vulnerability. World-writable single-namespace systems like Wikipedia work by having humans proactively checking for bad changes, and also by it not being that bad if a page is briefly defaced, since you can't push malware to users' machines that way.)

DoctorOetker• 1 month ago

there have been multiple package list checkers released, but unclear which ones are trustworthy, I can't immediately find a list from the official channels (through AUR website) to such approved scanners?

diogenes_atx3• 1 month ago

There are automated scripts to check if the infected AUR packages are installed on your system:

https://forum.endeavouros.com/t/malicious-aur-checkup-script...

https://github.com/lenucksi/aur-malware-check

jzer0cool• 30 days ago

What is a good review? Or different tiers of review? When thinking about packages, say in npm world, I still wonder about trust within nested dependencies. New users might also be unfamiliar. Any thoughts or any guidances?

AURpocalypse now: a look at the recent AUR attacks

Related Stories