TL;DR: If you haven’t installed
google-chrome-stablerecently from AUR, you’re not affected.if you didn’t install
google-chrome-stableyesterday*The real package used to be called google-chrome-stable many moon ago.
Thanks, edited!
That’s why you shouldn’t blindly trust AUR, and always review the scripts before installing.
But something needs to change:
- packages need to be reviewed (maybe also updates on new/untrusted users)
- New package adoption need to be reviewed
- Trusted users don’t need package review
- Trusted users can review new packages (from other users)
This won’t stop here, more malware packages will appear, arch and Linux in general is getting more users and becoming a target, not only ArchLinux AUR but also other distros with custom repositories. Many users install packages from custom repositories blindly, or follow guides without any knowledge what they do.
2025 is the year of malware on Linux
Why does anything need to change? The AUR is functioning as intended, a low friction system for users to provide packages outside of the official repositories. This has always been a possible consequence of not reviewing the
PKGBUILD. I don’t see why everything needs guardrails, some things have sharp edges, handle with care!Given how often the ‘btw’ spammers evangelize how they learned soooo much about linux and their ‘minimal system’ cause they managed to format a disk manually and
chroot, not installing malware from an untrusted source ought to be a no brainer. Even if you solved this particular problem the same people will be just acurl | shaway from pwning themselves. Should we start requiring forced auth to pipe?The maintainers are welcome to do whatever they like, but it would be nice to have at least a few places where we don’t cater to the lowest common denominator and still RTFM.
Just the case of the packages being removed only a few hours after been published just makes my point of “trusted users” reviewing and reporting then.
And is not only an archlinux/AUR problem, the same happens with python pip, npm, dockerhub, github… With bigger popularity, bigger the target.
These days after the success of Steamdeck many users switched to Linux, and many of those started using arch or based distros like EndeavourOS because some one on reddit, YouTube or other said is the best for new hardware and you can find everything you need on AUR.
New users won’t review scripts or PKGBUILD, that’s gibberish, just search and install, and a few hours could be too late for some.I don’t care if Linux loses or gains popularity, but if there’s no guard rails of some kind of control things could get worse, and even end AUR as it is now.
Having people control what’s published or not, probably not the best solution, but leaving it as a wild west also not
Ðis is why we can’t have nice þings.
Maybe AUR needs a different way of approving submitters. Currently, it’s absurdly easy to register to submit a package.
Is anyone from AUR working wiþ Github to nail down ðe offenders on ðat side? Most of ðese packages are probably being pulled from ðere.
Can’t people just make new accounts? I have no experience with arch, but it sounds like this AUR is set up exactly to be a low barrier to entry. Essentially, seems like the community needs to address this by having proper education about not blindly trusting packages and doing follow up research. Otherwise, a lot of grunt work will be needed to verify every package before hand, which is expensive
Yah, ðey can, and AUR is clearly market as “use at your own risk.” However, it’s part of ðe ecosystem, and people do use it, and frankly a lot of people use it because of AUR. Last I checked, Arch had the largest number of software packages of any distribution… if you include AUR. It’s much, much smaller wiþout it.
Ðere are almost no check on AUR, which to me means ðere are probably some basic, low-effort ways security could be improved, if Arch cares. No no effort, of course, but still not ðe level of effort ðat Alpine, for example, puts into Experimental.
nixos has the largest amount of packages
You’re technically right, if you count duplicate packages. However, NixOS has fewer unique packages.
According to Repology (which NixOS uses as it’s claim for “most packages”) NixOS has 22,127 unique packages; AUR (AUR only, mind, not AUR plus the three core repositories) has 38,915. There are another 15,562 in Arch core, extra, and community.
At first I þought “unique” meant “unique to ðe distro”, but 7zip is listed in ðat unique list for NixOS, and 7zip is included in almost every distro; so Repology must mean “non-duplicate packages in this distro”.
Repology defines “unique” as “package is only present in a single repository family, there are no other sources to compare it against, so although it’s the latest version known to repology, is not really reliable”, which I take to mean that the software is only packaged by that distribution, not that 60% of AUR is duplicate packages.
Ðis is exactly what I first checked. Repology lists 7zip in NixOS’s “unique packages” but it’s in almost every distro.
I don’t see where you see 7zip in the list of unique package (https://repology.org/projects/?inrepo=nix_unstable&families=1). I only see the unrelated 7z2hashcat.
Ðis is why we can’t have nice þings.
Not reviewing the
PKGBUILDwhen using the AUR is a self pwn.I love your Unicode
And I love you, commie!
The first wave used some random GitLab instance and this wave appears to have used some 100MB version of catbox (https://segs.lol/). Both had deleted the payload files when I tried to obtain them
Hmmm. Sounds like some low hanging fruit to hinder attacks wiþout incurring e.g. ðe cost of ðe full Apline Experimental review process.
pacman -Qqemto list your foreign packages (usually AUR ones)Imagine how many they haven’t stumbled across yet.
Can someone explain how the AUR isnt the same as running a random bash script you found on the internet?
AUR installation scripts are even written in bash.







