djad04
9 hours ago
I'm the one who wrote this.I originally tried submitting it here myself but without much luck so glad to see someone else shared it! Thanks for bringing it here.
also if anyone has another working PoC for that specific build I'd really appreciate if they could share it because looking at how the exploit works it does something that is not done on any other solution that i found for other builds, none of them hold the spot with a resize (key “1050” is sent twice) than make an intermediate buffer from the next field("1999") fall there.it’s the only approach I’ve seen that uses a repeated key like that.
I wonder whether there's a more direct way to achieve that as in the solutions for other builds. If there is, I think it would be interesting to compare the reliability of the different approaches based on the memory manipulation itself rather than how they reach the target
mhitza
an hour ago
I'm not sure you're going to see different outcomes with discussion this time around.
It's way too much text, branching out to explain too many things, has LLM-ism flavours to many sentences, and the AI generated images are not additive to your content.
For me to read it, it would need to be very much trimmed down in size (at most to 25% of current size) and punctual on what you've done instead of all the things that sidetrack the article.
exmadscientist
17 minutes ago
I agree that it's way too long and difficult to find the value in. One of the key features of a lot of good longer-form informational writing is that it goes around a subject before it goes into a subject. This takes many forms (and is, indeed, kind of fractal), but on its most basic level it's the "this is what I'm going to tell you, here let me tell you, now this is what I told you" pattern. And I say "fractal" because it applies not just to the whole thing, but to every chapter, and so forth. Even paragraphs often bear this structure.
When I don't know what you're going to tell me, I lose interest. When I hit descriptions of memory freeing right away, I don't know: are we going to blather in the weeds about pointers or malloc or ... ? When am I going to get to the part that's new to me? Etc. It's just hard to read, if I don't have to read it.
LLMs make this a whole lot worse because they don't need context like human readers and authors do. They can just jump right in. So they don't care, and they won't flag this, and if you end up writing for them, your writing won't be valuable to humans who need to learn.
jtreminio
28 minutes ago
I agree with the cutesy images, but for the rest ... it's called a deep dive, and it delivers. 25% of it might still have enough context about the bug and exploit, but would lose on the journey the OP went through to get to that point.