nr378

Joined 3 years ago

Karma

964

Submissions

139

Member since

Feb 15

Cambridge Computer Scientist

View on Hacker News

Recent submissions

Comment · 9 days ago

Whether it's exactly the same issue or not, the point remains: a double-free in and of itself doesn't establish a critical CVE (or even a high one). There's a known double-free in that GitHub issue, if there's a second independent one, it shouldn't necessarily cause any more concern than the first one.

Comment · 10 days ago

That post appears to show the same stack trace as this reported issue https://github.com/php/php-src/issues/24488, which is a confirmed GC regression in 8.5.12 but it is not exploitable (the allocator's shadow protection kills it), and it does not result in RCE and is not a critical CVE.

It's a very different shape to the other two CVEs, and a use-after-free doesn't necessarily immediately imply a critical CVE.

Comment · 10 days ago

That partial stack trace linked above seems to match this GitHub issue, which has much more information: https://github.com/php/php-src/issues/24488 (double-free introduced in 8.5.12)

From what I can see, this causes a crash only, and isn't weaponisable. It's a use-after-free, but the reclaim window is zero by construction.

Comment · 10 days ago

It seems like OpenAI are trying to frame this as if AI did all the work here, but it's not really true given what Calif say (https://calif.io/php2shell/):

> How did we find this?

> No special harness, or access to models that weren't generally available to everyone were used in the initial bug discovery. I focused on something I had a background in, knew its vulnerability history well, and held a theory that specific bugs lurk within the codebase. This shows that "know your target" is still as applicable as ever, even in the LLM dominated VR age. No model is simply capable of unearthing "all bugs" with simple prompts. At least for now.

> There are some practical research implications here. LLMs may have ingested public historical vuln data, but in their default variant analysis tends to move from the newest and most prominent bugs toward nearby older code. They do not reliably begin with obscure twenty-year-old exploitation primitives. Keeping your own mental catalog of a target's vulnerability history lets you prompt that direction explicitly.