Brotli vulnerability affects curl and Git

Posted by bmitch3020 4 days ago

Counter37Comment14OpenOriginal

Comments

Comment by bmitch3020 4 days ago

CVE-2026-111497 is not limited to nodejs. The brotli decoder is also included in other projects, including curl[0] and git[1].

[0] https://infosec.exchange/@harrysintonen/117489564359709891

[1] https://infosec.exchange/@harrysintonen/117489801970908776

Comment by altairprime 4 days ago

Elsewhere in this thread, they note that anyone decoding brotli responses is vulnerable, which would impact Apache/nginx conditionally; I can think of load balancer / proxy scenarios where that’s the case, as well as dynamic IP blocklisting as some use in-server, so duly noted for HN sysadmins.

Comment by Bender 4 days ago

The nginx brotli module compresses outbound traffic. How would someone exploit brotli on nginx in this scenario? Is there PoC code that could be used against an nginx server that has the Brotli module enabled?

Comment by altairprime 3 days ago

If you combine nginx brotli with nginx reverse proxy to a backend server that uses brotli-compressed responses, then you would have an opportunity for attack through this mechanism, assuming you could get your payload through both layers and back.

Comment by cowlevel 3 days ago

Just compress your POST request payload with it. The endpoint doesn't even have to accept POST requests

Comment by altairprime 2 days ago

Sure, that’s fine too

Comment by TFNA 4 days ago

Brotli is packaged along with MuPDF source code, too, and built as part of the MuPDF package by default.

Comment by TacticalCoder 4 days ago

> A buffer overflow vulnerability in the Brotli decoder (c/dec/decode.c)

It sure sounds like lately it's not a great time to be a C lover with a deep hatred on Rust...

And as a sidenote such exploits based on bugs in decoders is one of the main reason projects don't want to add support for every format under the sun (for example potential exploits has been invoked as a reason to not adopt jpeg-xl in a timely manner).

But yeah: maybe, just maybe, that C afficionados could concede that a language making buffer overflows impossible is actually a good thing? Instead of resorting to logical fallacies which are basically: "There are exploits for any single language out there, hence every language is equally likely to have buffer overflows".

Comment by cedws 2 days ago

I don’t get why Anthropic hasn’t already vibe ported all of this critical C code to Rust. If they claim LLMs can autonomously do ‘good’ language ports like they did with Bun, there’s pretty much nothing better they could do for the world’s security posture than eliminate dangerous C code. Why even have these cyber programs and then not do the most obvious beneficial thing possible?

Comment by cowlevel 3 days ago

We should demand them all to be written in WUFFS (not a joke)

Comment by cookiengineer 2 days ago

Remember yesterday, when all the top comments here claimed that "no real software is affected by this"? The same folks who wrote that the PHP RCE isn't real and downvoted everyone claiming otherwise?

Where are all those folks now?

For reference (the thread I am talking about): https://news.ycombinator.com/item?id=50219177

Comment by hulitu 1 day ago

> Remember yesterday, when all the top comments here claimed that "no real software is affected by this"?

There are a lot of Google employees on HN and their salaries depend on it.

Comment by rasz 4 days ago

>decode.c

What was that meme? "No Way to Prevent This, Says Only x Where This Regularly Happens" and from Google no less. Rare case of Mozilla W demanding JPEG XL decoder not be in C.