[PSA] Malware distributed on the AUR
On the 16th of July, at around 8pm UTC+2, a malicious AUR package was
uploaded to the AUR. Two other malicious packages were uploaded by the
same user a few hours later. These packages were installing a script
coming from the same GitHub repository that was identified as a Remote
Access Trojan (RAT).The affected malicious packages are:
- librewolf-fix-bin
- firefox-patch-bin
- zen-browser-patched-bin
The Arch Linux team addressed the issue as soon as they became aware of
the situation. As of today, 18th of July, at around 6pm UTC+2, the
offending packages have been deleted from the AUR.We strongly encourage users that may have installed one of these
packages to remove them from their system and to take the necessary
measures in order to ensure they were not compromised.
Follow up
There are more packages with this malware found.
minecraft-crackedttf-ms-fonts-allvesktop-bin-patchedttf-all-ms-fonts
What to do
If you installed any of these packages, check your running processes for one named systemd-initd (this is the RAT).
The suspicious packages have a patch from this now-inaccessible Codeberg repo:
codeberg.org/arch_lover3/brows…The Arch maintainers have been informed of all this already and are investigating.
like this
Mordikan, Endymion_Mallorn and Oofnik like this.
Tundra
in reply to Xylight • • •DirkMcCallahan
in reply to Tundra • • •Mark
in reply to DirkMcCallahan • •Linux reshared this.
copygirl
in reply to Mark • • •The "Arch cult's" holy book, the ArchWiki, states the following pretty clearly:
Mention of one's use of the AUR for their needs doesn't need to come with a disclaimer.
People who don't read or don't use their brain are going to keep not doing so, regardless.
tehn00bi
in reply to copygirl • • •irotsoma
in reply to tehn00bi • • •AntennaArray
in reply to copygirl • • •Well, it's not like cults were known for actually following their holy books
like this
Mark likes this.
FuckBigTech347
in reply to copygirl • • •caseyweederman
in reply to Mark • • •muusemuuse
in reply to caseyweederman • • •Helix 🧬
in reply to muusemuuse • • •muusemuuse
in reply to Helix 🧬 • • •caseyweederman
in reply to muusemuuse • • •teawrecks
in reply to muusemuuse • • •teawrecks
in reply to Mark • • •Nibodhika
in reply to DirkMcCallahan • • •That is sound advice, the AUR is most definitely not a trusted source though. For the normal arch repos the people who put the stuff there are known, they work for the project, you're as likely to get malware from one of those as you are to read an article bashing gamespot in gamespot, the people in charge of putting the packages there are the ones with more vested interest in things working so they won't knowingly introduce malicious code (plus it's a handful of people who know each other by first name).
The AUR is a different story, because anyone can put stuff there it's very easy to have malicious code end up there. It doesn't happen that often because most of the time it's fairly obvious and it gets flagged straight away, plus if people start doing that people will migrate away from the AUR, so it's a high risk low reward situation. But as more and more people start to use Arch derivatives that come with the AUR enabled without understanding any of this it becomes a more rewarding thing to exploit.
Dudewitbow
in reply to Nibodhika • • •caseyweederman
in reply to Dudewitbow • • •juipeltje
in reply to caseyweederman • • •Fecundpossum
in reply to DirkMcCallahan • • •The AUR, key words “user repository” is a specific weak point. It doesn’t have the same level of oversight that the main arch repo has. Stick to main repos and verified flatpaks and it’s very unlikely that you’d ever be compromised.
Linux isn’t perfect, but it’s certainly better than windows where you just download executables willy nilly to install your software.
HaraldvonBlauzahn
in reply to Fecundpossum • • •BTW python's package index has roughly the same problem - but a far less technical, experienced and critical user base. NPM has this problem since years.
Expect these problems to rise with every percent more of new Linux users which never learned the difference between opening / viewing untrusted data, and running untrusted code, because Windows basically ignores this essential concept and Android tries to solve that with sandboxing each app.
NuXCOM_90Percent
in reply to DirkMcCallahan • • •Yeah. The I'm A Mac crowd had the same problem... god damn it, two or three decades ago.
As market share increases, platforms become a much bigger target for malware. And a lot of the "I don't need to run virus scans" crowds learn the hard way.
Its the same with open source. Obviously NOBODY around here would parrot this bullshit, but there is the idea that because something is FOSS it is safe. Code is only as safe as code review and there have been a few high profile cases of social engineering to get malicious code past even fairly rigorous review. Let alone "Well, that script is FOSS so somebody probably reviewed it" that we see so often.
like this
thirtyfold8625 likes this.
Mwa
in reply to Tundra • • •bacon_pdp
in reply to Tundra • • •Only for distributions which don’t do reproducible builds and require full and complete corresponding source code under an FSF approved license.
If you choose to download binary blobs, good fucking luck.
Elvith Ma'for
in reply to bacon_pdp • • •Mwa
in reply to Xylight • • •Voytrekk
in reply to Mwa • • •pyssla
in reply to Voytrekk • • •Sorry, but I fail to see this.
I suppose if you're accounting literally all independent distros, then you're probably right. However, if we'd be more realistic and compare it to other well-established independent distros^[I'm basically counting Alpine, Debian, Fedora, Gentoo, openSUSE, Slackware, Solus and Void. I didn't count Guix System and NixOS for how their 'repositories' are built different and therefore not easily comparable to the others.], then we notice that the vastness of the packages found in Arch's repository is
... Show more...Sorry, but I fail to see this.
I suppose if you're accounting literally all independent distros, then you're probably right. However, if we'd be more realistic and compare it to other well-established independent distros^[I'm basically counting Alpine, Debian, Fedora, Gentoo, openSUSE, Slackware, Solus and Void. I didn't count Guix System and NixOS for how their 'repositories' are built different and therefore not easily comparable to the others.], then we notice that the vastness of the packages found in Arch's repository is rather lackluster at the very least. Heck, by virtually all metrics, Arch together with its derivatives undoubtedly belong in the upper echelons of usage stats; only being second to the Debian-family of distros. IMO, however, the size of its repository absolutely doesn't reflect this; as it's only bigger than Slackware, Solus and Void. The inclusion of these smaller projects is arguably charitable on my side*. But to drive the point home very clearly: Arch's repository is smaller than Alpine's, Debian's, Fedora's, openSUSE's and Gentoo's with a ratio of (about) two to one (except for openSUSE).
DistroWatch.com: Put the fun back into computing. Use Linux, BSD.
distrowatch.comVoytrekk
in reply to pyssla • • •I don't know if raw package counts is the best comparison. Unlike say Fedora, Arch bundles everything related to a project in the same file. If you want Qt6-base on Arch, that is one package. If you want it on Fedora, it is going to have a lib, header, docs, and maybe a few other packages.
Just from personal experience, I do not have issues with finding packages in the main repos, with only a handful of my packages coming from the AUR. This is not the case with others, like Fedora where extra repos need to be added, like EPEL and RPM Fusion.
pyssla
in reply to Voytrekk • • •Thank you for the quick response!
You're probably right. Do you think we got anything better to go by?
... Show more...Can't comment on this. Though, the list of packages with qt6 in their name is considerably longer in Fedora. However, I wonder if this simply reflects that Fedora, by virtue of having a larger repository, also has more stuff related to qt6. Or, as you posited it, chooses to package the same content over multiple packages instead of bundling them like it's supposedly happening on Arch.
Thank you for the quick response!
You're probably right. Do you think we got anything better to go by?
Can't comment on this. Though, the list of packages with qt6 in their name is considerably longer in Fedora. However, I wonder if this simply reflects that Fedora, by virtue of having a larger repository, also has more stuff related to qt6. Or, as you posited it, chooses to package the same content over multiple packages instead of bundling them like it's supposedly happening on Arch.
Hmm..., I feel you might be conflating stuff. Please allow me to elaborate on what I mean.
Fedora is not able to include some packages in its own repository due to legal reasons. As such, these are relayed to RPM Fusion instead. Which means that a well-functioning Fedora installation (almost necessarily) desires to install some packages from RPM Fusion. So, RPM Fusion exists as a 'hack' of sorts to protect Fedora from legal charges and NOT because they're too lazy (or something) to ship those packages themselves. To be clear, RPM Fusion is accepted as a trusted third-party repository.
Arch, on the other hand, is rather lenient on what they can include in their repositories. Basically enabling them to package within their repositories all codecs and whatnot without them being visibly worried about the legal consequences of this ordeal.
To be honest, I don't know exactly where this discrepancy comes from. But I wouldn't be surprised if it's related to how Arch is basically a genuine community distro while Fedora has official ties to Red Hat.
Btw, small correction, AFAIK you're not supposed to install packages from the EPEL on Fedora. Perhaps you meant COPR (basically Fedora's AUR) or Terra instead?
Getting started with EPEL
Fedora DocsMwa
in reply to Voytrekk • • •facow [he/him, any]
in reply to Mwa • • •cole
Unknown parent • • •Ulrich
in reply to Xylight • • •So...did someone just like create a new package cloning these or did they somehow get into the "official" repository? Is there no attestation process?
forbiddenlake
in reply to Ulrich • • •Ulrich
in reply to forbiddenlake • • •tehn00bi
in reply to Ulrich • • •Ulrich
in reply to tehn00bi • • •HERRAX
in reply to Ulrich • • •jenesaisquoi
in reply to Ulrich • • •MentalEdge
in reply to Ulrich • • •MentalEdge
in reply to forbiddenlake • • •To be clear, when projects distribute their software via the aur, someone else can't just issue an update using their package name.
This person appended "fix" and "patched" to appear in searches next to legitimate packages, and seem worth installing instead.
Jolteon
in reply to Xylight • • •MentalEdge
in reply to Jolteon • • •Absolutely.
The Arch User Repository is a way for anyone to easily distribite software.
Hence it has never been secure, and rather than claim it is, you mostly see people and documentation warn you about this, and to be careful if using it.
Any schmuck can make whatever they want available via the AUR. That's how even the tiniest niche project can often be installed via the AUR. But you trade in some security for that convenience.
Derpgon
in reply to MentalEdge • • •It shouldn't be used as a marketplace, it should be used as a repository. You can probably find a lot of malware on GitHub, doesn't mean you go there to choose your text editor.
I never search the AUR directly, I only use it if some README tells me I can install their software via an AUR package.
Dima
in reply to Derpgon • • •People need to remember it's not some carefully vetted app store and that they need to be the ones vetting any packages they install and any changes when updating.
aksdb
Unknown parent • • •HaraldvonBlauzahn
in reply to Xylight • • •What a nice attack on privacy-friendly infrastructure.
And then, Arch AUR has such suspicious things like the Brave browser which claims to reduce tracking.... and works together with advertisers.
To be clear, AUR is fantastic if you develop some experimental package and you want to give it to your friends to try it out easily. But not as a general distribution mechanism.
HaraldvonBlauzahn
in reply to Xylight • • •curl | bashinstall procedure and relying on TLS certificates which are e.g. issued by the Russian government. (No, the rust project won't use a Russian/Chinese/US Gov certificate but your browser will trust near all of them...)Ephera
in reply to HaraldvonBlauzahn • • •JackbyDev
in reply to Ephera • • •HaraldvonBlauzahn
in reply to Ephera • • •wewbull
in reply to HaraldvonBlauzahn • • •Ephera
in reply to HaraldvonBlauzahn • • •Sure, I guess, if you've got a distro installed on your PC and use the distro-provided packages to install the Rust compiler, then you can't be subject to such certificate MitM attacks.
Your comment sounded like you were primarily concerned about the shell script piping rather it just being a program which can be downloaded without going through distro packages.
HaraldvonBlauzahn
in reply to Ephera • • •The AUR install scripts are just downloaded shell scripts which are executed (hopefully after inspection).
curl | bashjust skips the inspection step - curl downloads to stdout, bash executes from stdin.muusemuuse
in reply to Xylight • • •We are getting to the point where inviting more people in means we will need an automated babysitter to watch for this shit and to pull it once it’s discovered. Apple has a walled garden approach that’s certainly taken a big chunk of malware threats out of their devices but their walled garden approach is ridiculous and impractical for Linux. The Microsoft method of monitoring and second guessing everything with antimalware programs is also suspect because it is super easy to abuse and resource intensive. We have clamAV but clam kinda sucks.
Linux is at the point where we need something that audits what’s going in and automatically yanks it back out remotely if it’s found to be a problem. Things can only be added by the user, but the bot can remove them without interaction of the user.
I don’t see this happening though. Instead, I see this as more of a rust vs C thing all over again, where valid critiques are drowned out by “improve your skills bro.”
mazzilius_marsti
in reply to muusemuuse • • •lagoon8622
Unknown parent • • •daggermoon
in reply to lagoon8622 • • •lagoon8622
in reply to daggermoon • • •curlthings intosh. Or downloading randomexes on Windows etcjuipeltje
Unknown parent • • •oo1
in reply to Xylight • • •I already assumed aur was riddled with stuff like that.
Use a condom when fucking around in there.
AceFuzzLord
in reply to Xylight • • •Gotta assume that if any Arch users actually fell for that one, that they either let their kids use their device or they're generally not smart ( which absolutely goes against my stereotypical view of an arch user ).
pfr
in reply to AceFuzzLord • • •moseschrute
in reply to pfr • • •I had no idea that existed but I’ve just returned from r/unixporn. There are some sick setups. Also we all copy. My entire neovim config is copied and modified from a couple dozen setups I admired. Nothing wrong with copying things you like. Don’t gate keep Linux.
However… Minecraft cracked is pretty funny lol.
lattrommi
in reply to moseschrute • • •I agree that gatekeeping is no good and people should not do that.
However...
I do not feel that assuming all people copy, should be done either, in my opinion.
pfr
in reply to lattrommi • • •lattrommi
in reply to pfr • • •I don't know if there is a word for what I was trying to point out.
Like an opposite to gatekeeping, sort of.
I do not like when people use 'we', in ways that include people that it does not apply to. Lumping everyone together inaccurately into a group.
lagoon8622
Unknown parent • • •teawrecks
in reply to Xylight • • •mazzilius_marsti
in reply to Xylight • • •the firefox, zen browser and libre wolf packages are concerning. The ttf ms font too. Those are very normal apps and unless you pay attention to the package name when doing "pacman -Syu", you would fall for the malware.
If only we can compartmentalize all AUR packages. The download AUR sources iirc are already in something like $HOME/.paru. Installing is a different story, because these packages can put their executable all over the places: /usr/local/bin, $HOME/local/bin.
Cysio
in reply to mazzilius_marsti • • •at this point you'll be reinventing Flatpak
SayCyberOnceMore
in reply to mazzilius_marsti • • •With respect, you wouldn't install these by just doing an update, so
pacman -Syuis fine.You would have needed to install these manually, or a package that depended on them - both from AUR - so you'd also need to use
yay(etc) to install them.But - I totally agree with your points that tge names look innocent enough for someone to install those over other packages.
Always look at the AUR (website) at the package details - if it's new(ish) and has 0 or 1 votes, then be suspicious.
Matt
in reply to Xylight • • •TLauncher — Download Minecraft Launcher
TLauncherbalsoft
Unknown parent • • •