Saagar Jha

As someone just catching up on the xz drama I have to say GitHub nuking literally everything related to the repository is incredibly frustrating. Real loser energy right there
8 replies →
8 replies

Sahil 🐧

(replying to Saagar Jha)

@saagar
Indeed, there is a backup of that git repo here: hachyderm.io/@joeyh/1121819336

David

(replying to Sahil 🐧)

@sahil @saagar

The original is still at git.tukaani.org/?p=xz.git;a=lo and still has the offending files.

Saagar Jha

(replying to David)
@hcetamd @sahil Yep I grabbed it, thanks

Parade du Grotesque 💀

(replying to Saagar Jha)

@saagar

Ask, and ye shall receive:

git.rootprojects.org/root/xz


tbodt

(replying to Saagar Jha)

@saagar maybe this was the only way they could think of to take down the tarball on short notice?

Saagar Jha

(replying to tbodt)
@tbodt I feel like taking down the tarball was perhaps not the best action to take anyways

tbodt

(replying to Saagar Jha)

@saagar no yeah this does seem like a net negative overall


Neia masks 🍵🌙

(replying to Saagar Jha)

@saagar@federated.saagarjha.com Yeah, I'd have locked any changes to the repository and, if possible, put up a banner about the vulnerability. The banner might require code changes, though, and localization.


Brian Campbell

(replying to Saagar Jha)

@saagar I expect this is a "it's Friday evening, we want to preserve evidence and prevent any further issues from coming up over the weekend, let's take everything down and sort it out on Monday" kind of situation.

The actual git repo is still available at git.tukaani.org/ (which includes the payload in the test binary artifacts), and the backdoored tarballs which activate it can be found on the Internet Archive: web.archive.org/web/2024032918

Saagar Jha

(replying to Brian Campbell)
@unlambda I wish they just locked the repo

synlogic

(replying to Saagar Jha)

@saagar yeah I feel lucky to have taken screenshots when I still could. I did not git clone down anything though, haha

Saagar Jha

(replying to synlogic)
@synlogic Thankfully there are countless mirrors

rlonstein

(replying to Saagar Jha)

@saagar Don't worry, I'm sure none of it got slurped into Copilot...

Saagar Jha

(replying to rlonstein)
@rlonstein I would be very impressed if Copilot did anything useful with it

Jeremy List

(replying to Saagar Jha)

@saagar I'd also heard the trojan wasn't actually in the git repo but was in a separate patch that was being applied between the "git pull" and "tar" steps of the release process (specifically after running autoconf) - if true the repo they took down would have been clean anyway.

Saagar Jha

(replying to Jeremy List)
@jeremy_list The backdoor was definitely in the upstream git repo, but it profiled the system when configuring itself to decide whether to build the malicious code in or not

Saagar Jha

(replying to Saagar Jha)
Also (provisionally while I fully catch up) it seems like everyone who *was* online when this stuff was around just ran around telling each other about the backdoor instead of looking into it. Come on if you’re going to do that you can make better memes than linking that one xkcd

Saagar Jha

(replying to Saagar Jha)
Report: This injects an obfuscated script
Me [reading the Makefile changes]: Ah yeah I can see how that is pretty obfuscated and difficult to understand
Report: …I haven’t even gotten to the obfuscated script yet
Me: Don’t bother this is too hard for me to read already

Saagar Jha

(replying to Saagar Jha)
I need to apologize for an earlier tweet where I said people weren’t looking at this. After looking at this code I have to say “I am *not* a security researcher, nor a reverse engineer” is the understatement of the year. The malware is quite sophisticated and difficult to analyze
1 replies →
1 replies

Khaos Tian

(replying to Saagar Jha)

@saagar we got really lucky this time… in another timeline it’s easy to imagine this get pushed to the stable Debian releases and exists years without getting noticed 😅

Saagar Jha

(replying to Saagar Jha)
It would make for a decent CTF challenge–I’m surprised someone without reversing experience was able to get as far as they did. There are people actual security people in my timeline who did put some effort but the reason I didn’t hear about it is that they didn’t get very far
1 replies →
1 replies

Dominic Hopton

(replying to Saagar Jha)

@saagar wait till it turns out this was just one of:
1. A precursor to a new interview take home test for a security company (“Find the vuln”)
2. Someone doing a take home test for a security company (“Insert the vuln”)

Saagar Jha

(replying to Dominic Hopton)
@grork People get upset about one hour take-homes and this company is making people do two year tests…

Dominic Hopton

(replying to Saagar Jha)

@saagar they only hire the best of the best

Saagar Jha

(replying to Saagar Jha)
Definitely not impossible to fully analyze of course but this is way past your “lol click around in IDA” stuff

(Full disclosure I poked at it with that level of commitment myself because was curious and very quickly abandoned the effort as too much work)
3 replies →
3 replies

Sahil 🐧

(replying to Saagar Jha)

@saagar
IDA costs a fortune!

Saagar Jha

(replying to Sahil 🐧)
@sahil Yeah I actually mostly use Binary Ninja for personal work (including this)

Sahil 🐧

(replying to Saagar Jha)

@saagar
Yeah, I was trying on binary ninja too yesterday but couldn't find anything XD

My attempt on Binary Ninja

Saagar Jha

(replying to Sahil 🐧)
@sahil This is one of those things where the tool matters much less than the person looking at it and the effort they put in

osy

(replying to Saagar Jha)

@saagar wasn't your full time job for almost a whole year to look at Android malware samples for Google

Saagar Jha

(replying to osy)
@osy86 No my job was (…checks half-updated résumé) “developing resilient techniques to thwart bad actors targeting Android by following them into where they operate”

Sahil 🐧

(replying to Saagar Jha)

@saagar
Someone reversed engineered it. It's RCE.
Source: bsky.app/profile/filippo.abyss