AMD is warning about a high-severity CPU vulnerability named SinkClose that impacts multiple generations of its EPYC, Ryzen, and Threadripper processors. The vulnerability allows attackers with Kernel-level (Ring 0) privileges to gain Ring -2 privileges and install malware that becomes nearly undetectable.
Tracked as CVE-2023-31315 and rated of high severity (CVSS score: 7.5), the flaw was discovered by IOActive Enrique Nissim and Krzysztof Okupski, who named privilege elevation attack ‘Sinkclose.’
Full details about the attack will be presented by the researchers at tomorrow in a DefCon talk titled “AMD Sinkclose: Universal Ring-2 Privilege Escalation.”
I’d like to see the people that are for kernel-level anticheat running in their PCs 24/7 now.
Vanguard doesn’t even let you play the game unless the anticheat has been active since boot afaik.
Every cybersec and even anyone minimally tech-savvy was saying this was a bad idea and what do you know, now we have objective evidence it is in fact a terrible, terrible idea.
No fix planned for Ryzen 3000 is a disappointment
That is indeed weird. I don’t see a mention of summit and pinnacle ridge either, even though rome, castle peak and renoir are covered. Since this is fix is an AGESA upgrade, I’m inclined to believe it’ll cover zen 2 on any platform.
I’ll reach out for clarification. I have a Matisse X production system running and there’s no way I’m leaving this be.
How it that conceivable?
Not that it changes anything, but it might be worth to voice our concerns
I found this thread on their forum:
Yeah it really is, but a bit of a nudge to find a decent price on a 5600x.
For everyone who’s posting but didn’t actually read the CVE…
You need Malware with Kernel level access Already. Besides Anti Cheat for modern games, if you have malware with Kernel level access you’re already really fucked.
In addition, this just appears to be a way for that Kernel malware to persist in a device. It’s not impossible to detect. I wouldn’t be surprised if we saw Windows Defender signatures for malware using it within the month.
I haven’t seen the Def Con presentation, but the CVE is a “Maybe”. There is no PoC (Proof of Concept, showing that an exploit works)…yet. keep an eye on that.
The CVE claims this “Could Maybe” allow based on logic, but none of the sources i found showed anyone actually using it. Maybe the Def Con presentation will. But unless I see someone post a repeatable exploit in a real world scenario, it feels superficial.
I want to reiterate that this IS a flaw and it IS a problem. But I would highly doubt you, rando consumer, will be affected.
Mitigations are to not be dumb on the Internet. Keep your browser updated and make sure your sensitive data is backed up and encrypted, basic stuff.
Happy to go in to it more. This is my jam.
Maybe we could use this to disable the Intel ME equivalent for AMD on those generations.
PSP is required for x86 core initialisation.
But if you can get around the signing requirements, it would be trivial to make the psp firmware only do that. (Probably not trivial, but at least technically feasable)
I did want to expand on this, I’d be curious as to what Jeremy Soller found getting certain Ryzen mobile platforms working with coreboot.
Perhaps we’ll need to wait for OpenSIL for more comprehensive support.
Aside from x86 core init, PSP is still required for specific OS security requirements. Can at least take solace in the fact that OEMs seem to have completely turned down MS Pluton.
The disclosure has been updated to include a fix designated for Matisse
https://www.amd.com/en/resources/product-security/bulletin/amd-sb-7014.html
Looks like it’s time to upgrade…
Iiuc to exploit you already need kernel level access. I’d figure if something malicious made it that far, you’re already screwed.
