top | item 46326120

XZ Utils Backdoor

25 points| ctrlmeta | 2 months ago |en.wikipedia.org

21 comments

order

LunaSea|2 months ago

> The Debian development team declined to remove the affected images, stating that they were development builds that should not be used on real systems in place of newer, clean container versions.

Classic Debian security management

BrouteMinou|2 months ago

Not that I approve the Debian decision here, but calling it "classic" seems a bit of a stretch?

Do you have many more examples to call that a "classic" Debian security behaviour?

jmclnx|2 months ago

>While xz is commonly present in most Linux distributions,

Not Slackware since Slackware does not patch xz or many other utilities. Plus it does not use systemd. From what I remember a patch was put in to give systemd extra functionality and someone used that patch to sneak in the backdoor.

NekkoDroid|2 months ago

The xz attach happened cuz systemd's library dynamically linked against xz for compression of various tools in systemd and a downstream patch for openssh (IIRC) was used to link against libsystemd to use some founctions for the sd-notify protocol.

This wasn't exactly necessary since the protocol has been stable for external use for ages (since its inception IIRC) and is relatively trivial to implement.

Since the attack happened openssh gained native support for the sd-notify protocol, the sd-notify man page has an example implementation that is freely usable and libsystemd now only loads xz (and most of its other libraries) when explicitly requested by one of the tools via `dlopen`.

flykespice|2 months ago

The behemoth that is autotools mostly helped to conceal the backdoor (and contributed to the payload)

It's an old legacy technology that needs to die out from all forms of distributions (looking at you GNU)

pogopop77|2 months ago

Given this was backdoor was likely funded by a nation-state actor and very carefully obfuscated, the fact that it was discovered within a month and never rolled out to production releases, shows that the open source process mostly worked. Not saying it couldn't be better.

mingus88|2 months ago

I kinda disagree. This was luck. A dev on an unrelated project happened upon it and was diligent enough to dig in. A single change to any number of variables would have meant disaster.

I worked at a company that got red teamed. The pen testers were inside the network and were only found by a random employee who happened to be running little snitch and got a weird pop-up

Nobody celebrated the fact that the intrusion was detected. It was pure luck, too late, and the entire infosec leadership was fired as a result.

Like this xv issue, none of the usual systems meant to detect this attack seemed to work, and it was only due to the diligence of a single person unrelated to the project was it not a complete show.

jqpabc123|2 months ago

This really illustrates a broad security issue with open source development and methodology.

Who vets contributors, maintainers and submissions?

Answer: Unknown in many (if not most) cases. Unless you have the time and expertise to do so yourself; it is purely based on trust.

nwellnhof|2 months ago

Contributors and their submissions are vetted by maintainers. New maintainers are ideally vetted by existing maintainers. This can obviously break down in undermaintained projects.

yjftsjthsd-h|2 months ago

That's not unique to open source or open development.