If you’re confused why you can’t currently download Ubuntu 23.10 despite the fact it’s been released (and blogs like mine are telling you it’s out) there is a reason.
[From Twitter]: “We have identified hate speech from a malicious contributor in some of our translations submitted as part of a third party tool outside of the Ubuntu Archive. The Ubuntu 23.10 image has been taken down and a new version will be available once the correct translations have been restored.”
Now, I’m not 100% certain but from poking around the Ubuntu Desktop Installer GitHub — I know, I’m nosey — appears to have been (sadly) the Ukrainian translation file that was hijacked. I ran the text through a translator and …Honestly, I wish I hadn’t.
It’s a broad range of offensive sentences touching on politics, sexuality, and current events. Though shocking, none of it is particularly coherent in scope. It seems to be written to be provocative for provocations sake – the sort of stuff people post on X to farm likes from far-right bots.
Nobody is even slightly concerned that this made it to release? if they can shove in hate speech without anyone noticing, cant be much harder to slowly introduce a backdoor over several commits.
Minecraft got in trouble when the Afrikaans translation had the n-word (in English) due to a malicious translator. CDPR had an issue with the Ukrainian translation making references to the ongoing war.
This sort of thing happens somewhat frequently. It’s the same reason how fake sign language interpreters can hold positions. It’s hard to verify the accuracy of a translation in a language you don’t speak. They have to trust that the translator did their job right.
Translations are usually just text strings. No reasonable project would allow translators to write code.
I mean honestly though, if there are code reviews, how hard would it be to just make a quick “translation review”, putting the stuff through a translator program, and verifying it’s not obvious bullshit? Especially for new/unknown contributors. Of course it’s additional work, again, but a sanity check should easily be possible.
Quite hard. We had Open Source’ish LLMs for only around six months, if they are even up to the task of verifying a translation is another issue and if they are up to Debian’s Open Source guidelines yet another. This is obviously going to be the long term solution, but the tech for that has simply not been around for very long.
And of course once you have translation tools good enough for the task, you might just skip the human translator altogether and just use machine translations.
I more meant that if something contains “fucking kill all ukrainians and trans people”, which it sounds like this was something like that, that should be possible to see even with bad translation tools.
It wasn’t, by the way. Though it could have been flagged by the dumbest of online translators (or even anyone who could read Cyrillic, since some of it uses English loanwords, like “sex” and “gay”). It should never have made it in release, but I disagree with categorizing it as “hate speech”. I feel comfortable posting it here, even though it’s pretty crude and #3 in particular is very vulgar. If anyone’s curious, here are the Google Translate translations of the vandalized parts (except for one of them, fullInstallationSubtitle, which I think is too offensive to be repeated here. It references the Israel-Palestine war):
I mean yeah, I was speculating. But what you posted also seems easily detectable :D
That’s a lot of pant removal.
But anyway, it should be quite possible to automatically screen the translation for something this blatant.
I would assume since it was a block of raw text in Ukrainian in a translation file, it would have passed more under the radar than something like a backdoor. I do not know how things are reviewed before being pushed to release though.
Not really, not only because of the language but also because the same scrutiny between code and content wouldn’t have to be the same. I also don’t expect core aspects of the distribution, e.g kernel, package manager, cryptography libraries, to be verified the same way than a random software, e.g Kdenlive. So… is it bad, absolutely. Does it mean everything should be questioned again? Probably not.
I’m sure more people know C or Python than Ukrainian at Canonical. It looks like this particular change has been authorized by a third-party localization project, though I’m not sure the whole process works.
Translations are not going to be analyzed as thoroughly as code, and this was still found quite quickly. Submitted code is analyzed much more thoroughly, often by multiple members or the project.
It is very concerning, absolutely. With that said, it’s entirely possible localization/translation reviews work differently than code reviews.
Well but they DID notice
Most translations are contributed by external users for languages that the project developers don’t speak themselves, so they can’t always check everything unless there’s multiple active translators for one language.
Ukrainian has enough speakers for there to be multiple translators, doesn’t it?
Clearly not enough active ones for each and every project out there.
But oPeN sOuRce iS sAfe.
Lol. You have to understand the context here. This is just translations. Actual code has many, many more eyes on it. An entire university was banned from submitting code to Linux, because of two dumbasses. They found and fixed genuine bugs. Built up lots of trust. Then violated that trust with actual use-after-free bugs submitted intentionally.
The submitted “patches” to the development branch was to prove it’s easy to get exploits into high profile open source projects. They ultimately proved the contrary. Making their “research” bunk. The code they submitted never made it past the development testing phase.
The context is that code made its way into shipped open source software.
The type doesn’t matter. It proves that there can be slip ups.
Move goal posts, though.
Something nobody has ever disputed.
Translations… aren’t code though…?