It certainly does suck. Perhaps a silver lining is that I’ve adjusted my use case … minimizing AUR packages to the greatest extent practical. Maybe even more as time marches on.
The key takeaways here, which would apply to all the earlier issues and this new flag, are:
beware if you use a package that is recently adopted
beware if you use a newly created package
That doesn’t mean don’t use it, because it may well have been adopted or created by a trusted maintainer, perhaps even one of the Arch named “maintainers” (essentially admins). It simply means, beware, and pay attention to all changes made to that package.
And to be clear, I’m not saying that you can just assume everything else is completely safe. Only that recently adopted, and newly created packages are where the key exploit opportunities have been.
Why target AUR? Isn’t Arch a niche or a small subset of linux base? Even servers are not based on Arch but rather other distros, like Debian, RedHat/Fedora, openSuse/suse, etc.
If one looks at the packages it seems that the packages which were supposed to be targeted for adoption are all developer or android focused. Are there some frontier AI labs or other critical significant economic actors which are using Arch or one of its derivatives?
It does not make sense to go after such a small user base, by trying to target AUR. I mean what is the motive over here?
I think if anyone has managed web-facing assets for any significant amount of time, they’d be aware that bots and opportunistic actors poke and prod your assets at a constant pace, always probing for that vulnerability that’ll help them open the next door.
Targeting developers is not surprising, because developers will often hold the keys to additional more meaningful assets on their systems.
So basically compromise Arch systems and then make a jump from those systems to the intended target. Just like the OpenAI agent(s) did a few weeks ago.
Would that be a correct characterization?
Because being honest and not being disparaging, our total user base is small compared to other distros.
You know CachyOS? That distro is used even by complete newbies because it was spammed everywhere, despite the fact the distro itself is based on Arch Linux. Arch based distros also top Steam charts, alongside Arch itself. You might think “but that’s only gamers”, yeah, you’re right, but it’s clear that Arch Linux based distros are quite popular among Linux users.
Why not other distros? Well… as far as I know:
Debian doesn’t have anything like the AUR;
Fedora has RPM Fusion, but those are peer-reviewed and it acts basically like an official extension of the official repositories;
openSUSE has Open Build Service, which acts more like a centralized build server more than anything. You can absolutely inject malware here because it also works on a trust system similar to the AUR, but not many people use openSUSE and in a server environment with SLES, sysadmins probably don’t want to risk it.
I don’t think so. I think it’s just a bunch of clowns just messing with the AUR because they can. Some packages aren’t exactly developer focused. osu-lazer-appimage was one that jumped on me. That’s a game, not a development tool. Admittedly, the AUR has a lot of versions of the game available on it, for some stupid reason, so I am not entirely surprised one of them fell prey to this. But to me that makes it somewhat clear that they are just targeting any package they can. Seems like an automated thing as well.
The AUR, along the OBS from openSUSE, are perhaps the only user-generated and maintained repos that are actively pushed, to some degree, by the official maintainers of a distro. When the official website for your distro has a button to go straight to the user-generated repository RIGHT NEXT to the Download button, when the culture around the distro is quite heavily focused on using the AUR instead of channels that are more officially endorsed by the developers, meanwhile the repository’s security model is based on trust and reading diffs, I’m sorry to say this because I do like Arch Linux, but it was kind of bound to happen, especially now in the age where LLMs can, somewhat, do this type of BS at a moment’s notice automatically.
There’s a discussion here about security models, how to actually prevent such attacks/make them harder to perform, if the AUR even has a place in the modern landscape of packaging on Linux, etc. I’m not going to throw my hat in because I did and my take remains the same: just use whatever the developers tells you to use for your specific distro because that’s probably the most secure way to get whatever you want to get.
Going back to that osu! game as an example: on Linux, the game is distributed from the main website through an AppImage. I just use that. I manually made a .desktop file for it for easy access and I use an AUR package that simply just adds MIME definitions for the various file formats the game uses (all of which are just disguised .zip files). It works fine. It’s more annoying than downloading the setup program and running it like on Windows, I also use the AUR because it’s annoying to add MIME definitions yourself (I’ve done it before and it wasn’t too bad, but it was annoying), but it’s not too bad, I guess…
Much of the focus since the original wave of AUR attacks has been on the injection of binaries into previously well-behaved installation scripts…
Is there much info available about just what some of these nefarious binaries are actually designed to do, functionally?
The idea of being broken into is disturbing enough. I’m just wondering how malicious the intrusions have been thus far in this latest chapter of the security arms race?