Adoptions are a mistake anyways. Remove unmaintained packages and block the name for several months.
Remove unmaintained packages and block the name for several months.
Alright, but that would mean most of AUR. Have a look at the package statistics box on the AUR homepage. Most packages fall under that definition in one way or another. The vast majority of the AUR is package some random person added once then never bothered with ever again.
Frankly I’m surprised that the AUR has survived for so long in its current form, for what is basically a shell script distribution system with zero supervision and zero safety guards.
The vast majority of the AUR is package some random person added once then never bothered with ever again.
Yes, that sucks. But I was thinking of exclusively this packages: https://aur.archlinux.org/packages?submit=Orphans
The orphan package criteria is misleading. If you look at the infobox you’ll see that there are only 69 package “maintainers” for 100k+ packages.
THIS is misleading, though.
There are definitely more than just 69 users who uploaded packages to the AUR.
Adoption of unmaintained packages to maintain them is not a mistake. The problem is the current implementation, not the idea behind it. It’s like saying the AUR is a mistake, because some people do malicious stuff.
They should find a better solution, like adoption shouldn’t be granted to everyone without question, especially new accounts who didn’t maintain anything before. Mass adoption shouldn’t be granted automatically (limit rate), in example 1 package adoption per day and if someone wants more, admins or moderators need to approve. And updates of newly adopted packages should wait a day.
Also the AUR helpers should do a better job. Always ask if a new adopted package should be updated and give a warning the maintainer changed.
They should find a better solution
Who’s “they”? Because it’s not Arch. Arch doesn’t want to have anything to do with AUR, and neither does any of the Arch-derived distros. They’re all perfectly happy taking advantage of it, of course, but not the responsibility.
Who’s “they”? Because it’s not Arch. Arch doesn’t want to have anything to do with AUR, and neither does any of the Arch-derived distros. They’re all perfectly happy taking advantage of it, of course, but not the responsibility.
Where did you got this nonsense from? What do you mean “they are not Arch”? The AUR is managed and operated by the Archlinux team. As the packages are community-driven content, they cannot guarantee and give support, because it is not their package. But they are still managing and supporting the AUR itself.
- https://archlinux.org/news/active-aur-malicious-packages-incident/ from 2026-06-12 is an official message on the main Archlinux website (there is no new post about the current situation).
- from https://archlinux.org/people/package-maintainers/ : Campbell Jones is an AUR Moderator
- from https://archlinux.org/people/support-staff/ : Andrea Denisse Gómez-Martínez is an AUR Packager, Robin Candau is an AUR Moderator
And you’re gonna see the Arch team wash their hands of the whole thing, like they did in the past whenever the AUR was in trouble.
That’s not real ownership.
Do you have any sources, links or evidence for your statements?
Is the current state of the AUR, despite the previous waves of attacks, and its troubled history, not evidence enough? The Arch team has never made the AUR a priority and I don’t see why they would start now.
The way I see it there are three possibilities:
- They do nothing.
- They shut it down.
- They give it up for adoption.
What is not going to happen is the Arch team putting time and effort into overhauling the AUR.
I thought so.
The problem is the current implementation
Yes, exactly this! I am not surprised it happens. I’m surprised it didn’t happen before.
especially new accounts
Weren’t there “sleeper accounts” registered years ago that became active in the current wave?
Also the AUR helpers should do a better job.
Even experienced people will just update as if nothing could happen. Adopted packages should have to use a different name and the current name being blocked so it WILL get attention when some tries to update their system.
Weren’t there “sleeper accounts” registered years ago that became active in the current wave?
Which does not invalidate my point about new accounts, but yes. Its important for new accounts, so once a sleeper account is banned, its not that easy to just create thousands of new accounts while everyone is focusing on the current active ones. And if, as I suggested, mass adoption per account is not possible, then the attacker has less to attack.
Edit: They need to make sure that sudden editing many packages in short time, with probably the same or similar lines should be automatically reported. They need some automated checks in place, at the very least. This would be a very suspicious behavior if many accounts are not active, and then suddenly all of them do something.
Even experienced people will just update as if nothing could happen.
Then you can’t fault the system, if you are this reckless.
Adopted packages should have to use a different name and the current name being blocked so it WILL get attention when some tries to update their system.
This would break all dependencies of this package. The point of adoption is to keep it working and stable. I’m highly against force changing the name, that is not a solution (at least not one I am happy about).
[Changing a package’s name] would break all dependencies of this package.
Yes, that is correct. But that should not be a big deal for packages that are actively maintained. The maintainer can simply change the dependency to the new name after making sure the new package is legit.
OK, that’s a good point. It would prevent auto updating. However any package that is NOT updated, should stay with same name in my opinion. So that everything (with the old secure code) stays intact and working. The name change should be part of the the update process. So it only breaks if you want to update, which would ensure compatibility if you choose not to (as it is safe). Something along the likes like this. I agree with your suggestion now, because that seems to be sensible idea.
Enforced commit signing and making it obvious when the signature changes would help make adoption much safer. I really hope they implement it.
Just saw DeepSeek TUI in there, that’s probably going to be a whopper that hit quite a few users.
You mean https://aur.archlinux.org/packages/deepseek-tui-git from the list https://lists.archlinux.org/archives/list/aur-general@lists.archlinux.org/thread/P4WIRHTFNH2YZWQHGBAKQWX5YOAFIDLY/ ? It has 0 votes and Popularity is at 0.000000, since it was uploaded 2 months ago. And there is a non Git version that is not listed (probably not affected) at https://aur.archlinux.org/packages/deepseek-tui . On top of it, on their website https://deepseek-tui.org/#install the main recommended way to install it is with
npmand there is even a way to build it with Rust it seems (I’m a bit unsure).So I don’t think it hit that many users, as only a couple of people since Mai would want to install it, and they have to have updated it to build from source in recent days. That is a very narrow number of people, just by my gut feeling.
Tumbleweed stay winning
No way to prevent this, says only repo where this regularly happens
Isn’t this similar to the reason a lot of people hate snaps? Or am I misunderstanding something? I’m not an Arch user (btw) so I’m not super familiar with AUR.
You’re right on malware finding its way onto the Snap Store. I find it hilarious to see that the employed tactics are basically identical 😜.
However, FYI, the hate on Snaps is a lot more broad than that.
Not the only repo, see: npm
Npm doesn’t let you easily take over packages you don’t own.
but does let you take over their maintainers’ accounts and easily poison them
Not more easily than anything else.
Besides all the other non infected ways to install the software, there is a way to prevent this: Just read the AUR package before install and don’t trust blindly any new maintainer.
It’s metaphysical approach to security. Enshrined rules that can’t be enforced don’t define user’s behavior.
