Researchers discover potentially catastrophic exploit present in AMD chips for decades
engadget.com
Intel breathes a sigh of relief as the spotlight moves off of them for a beat.
Intel breathes a sigh of relief as the spotlight moves off of them for a beat.
Except the AMD exploit requires ring 0 access and is almost irrelevant to most users, whereas the Intel issues are physically destroying people's computers. The scale of the issues are utterly incomparable.
I'm much more angry with whatever dipshit at AMD decided to revoke permission for ZLUDA, and that they haven't yet been fired.
People grant kernel access all the time without thinking. Video game anti-cheat is a good example. It's a pretty potent vector of attack since you can never trust these companies to keep themselves secure.
Apex Legends and Genshin Impact being a good example of this.
The Apex Legends hacking situation was unrelated to the anti-cheat despite initial reports. It didn't stop the hack but it also wasn't the vector for the attacks.
Genshin Impact's anti-cheat however was an unmitigated disaster.
you're right, thanks for the clarification! https://inv.tux.pizza/watch?v=lzW4SDm0yWM
I'm thinking there are more things like this, and maybe even some are intentionally put in to please government actors. Intels management engine is like a small computer inside the computer, and amd has something simular. This computer can work completely independent of what the main computer does and has network access.
ZLUDA was probably taken down to protect AMD from being sued by Nvidia.
https://www.tomshardware.com/pc-components/gpus/nvidia-bans-using-translation-layers-for-cuda-software-to-run-on-other-chips-new-restriction-apparently-targets-zluda-and-some-chinese-gpu-makers
Didn't the author confirm the takedown came from AMD and not NVidia? AMD isn't responsible for third party software running on their hardware.
Although, IIRC they either sanctioned it or provided some initial funding, which might have put them in a more culpable position. Still, I'm pretty sure the takedown came from AMD, and it doesn't make sense that they're doing NVidia's policing for then.
They were paying for its development for about a year and a half.
Yeah, that'd do it. Although, again, it looks like the restriction wasn't in the NVIDIA licensing wording until recently. IANAL, but you it both parties are required to agree to contract changes; if AMD's contributions were all pre-wording change, they merely need to dust their hands; it's OSS. Why are they doing NVIDIA's dirty work for them?
I'm not convinced.
what's ZLUDA? can you ELI5?
ZLUDA was an open source translation layer for CUDA. So basically developers could take code from projects written for Nvidia's CUDA and use ZLUDA to run them on other hardware. Originally the dev was focused on Intel but AMD started paying him and he focused on AMD hardware. They stopped funding him earlier in the year and now it appears AMD legal has gone back on their earlier permission for him to keep distributing the code.
oof, what a rugpull
This post was sponsored by INTEL™️
/jk
How long were you sitting on that one waiting for the perfect moment?
Which is by pure chance also what the person said who got the exploit list from Intel
Wouldnt be surprised if the researchers were Intel researchers.
*100 employees were fired too allocate budget for this sponsorship
Funny. No. This is not remotely comparable to CPU's crashing because of a lack of R&D and half ass quarter earnings profiteering culture's lack of intelligent long term thinking. It is not remote accessible. This is just corporate psyops in an attempt to coverup their overwhelming neo feudal incompetence. If they had the staff and invested reasonably, the problems wouldn't happen. Paint the world in shit to continue the claim that yours does not stink. Only idiots buy into that.
Physical access is root access, always has been
"FUCK!" - Intel execs
Hmm.
It does mean that any secondhand computer or CPU (or even CPU from a sketchy source) could be compromised prior to being physically sold.
I have worried a bit before about the physical supply chain. Consider this case, earlier in the year, about someone selling counterfeit Cisco hardware (not intending to compromise computers, just make a buck):
https://www.justice.gov/opa/pr/leader-massive-scheme-traffic-fraudulent-and-counterfeit-cisco-networking-equipment
I remember that that hardware made it into even Cisco's own authorized partners' inventory.
And that's not something that's gonna be far up in the supply chain. People don't build Cisco hardware into a lot of other products.
So you gotta wonder what can happen if someone has a good way to undetectably compromise CPUs and insert them into the supply chain.
Unless I'm mistaken, the malware isn't on the CPU. The exploit is CPU, but the firmware is stored on the bios chip. Used motherboards are a potential for having malware on them, but then again they always have been a risk
It's worse than that, any AMD chip from any source except maybe AMD directly is suspect. Mine is a few years old from Amazon supposedly new, for all I know it came compromised and is sitting there doing what I tell it to until it triggers and I won't even know when or if it happens.
That's not how this exploit works at all...you have to have physical access to the machine basically. This is a nothing burger.
That's to get it installed, not if it's already there.
It's not going to be there because if you're compromised via physical access, no one is going to give a shit about this exploit.... it's like someone having the keys to your house and then being worried they're going to smash out a window to gain access.
I don't think you're following along here. The physical access would have already happened prior to the CPU even being in my possession.
I see what you're saying. You're assuming someone grabbed a bunch of cpus, fucked with them, then tossed them back into the box and sold them as new.
Exactly, if I were a bad actor with access to the stock that's what I'd do. I'm sure there are multiple points along the supply chain where it could potentially happen.
That's true, seems like you'd need to know where they're going though, like a ton of work just to hopefully get one machine infected that has anything on it.
If I understand it correctly, the chip has the vulnerability, but the malware would be installed on the motherboard in the form of a bootkit. So getting a used CPU is not a threat, but getting a used motherboard is (and kind of always has been) a risk.
It allows for adulteration of firmware, the CPU has firmware. 🤷
CPU firmware exploits are incredibly rare, if there even are any that exist beyond proof-of-concept. The chances of getting an infected CPU from this is so unlikely it’s practically impossible.
You forget that the CPU has a nanny CPU built in these days.
Which, again, is an incredibly unlikely attack vector unless you have some government secrets on your computer. And chances are that any attack through the IME or PSP is trying to do an implant into the UEFI/BIOS and not the processor itself.
Oh man, the Pro Networking scandal is so funny to me. I wonder how many of those machines are out there running IOS right now and no one is the wiser. AFAIK there aren't allegations of backdoors or anything, just fake Cisco gear.
Does it mean that? I mean a computer bought from a sketchy source, sure. But just a cpu alone? Do these raptor lake cpus have any non volatile memory? Because if not, then a second hand cpu is totally safe.
This is an attack vector that hadn't occurred to me and I find it disturbing. Wow.
It was really just a matter of time before someone figured a way to exploit those stupid deep management engines. It was so predictable.
And each intel chip runs a minix system behind the scenes that I'm sure someone will soon find a way to play with if it's not already compromised.
There was a big story of an expoit of the Intel ME already. I think it was this: https://www.intel.com/content/www/us/en/support/articles/000025619/software.html from Positive Technologies at BlackHat 2017
Sinkclose
Impacted systems: https://www.amd.com/en/resources/product-security/bulletin/amd-sb-7014.html
Original presentation details. https://ioactive.com/event/def-con-talk-amd-sinkclose-universal-ring-2-privilege-escalation/
This is a big f****** deal. If you get kernel access to a system, which is not uncommon, you can install a persistent back door that subverts the rest of the system forever. That's huge. It does not require physical access, requires kernel access.. different things
Steam games that ask for admin privileges, a USB stick that boots, normal virus infection that can never be uninstalled,. persistent RAT....
Negative rings are a horrible proprietary liability.
That's been clear from their inception, and this changes nothing.
What the heck? Then get to planning you dicks?
There is no possible fix.
For all the others on the page they are planning fixes, only Mattisse has this. Is there something special about Matisse compared to the others?
I heard hammers can fix things
I'll bet the Intel management engine is just as "vulnerable". The only context this is likely a concern is large scale corpo deployments, without verified supply chains to the source. Love how the security researcher handwaves that there's "plenty of existing exploits" that can be used to install the exploit into the SMM, without giving any suggestions of how.
No way that Intel sat on this for years until they needed it for PR.
I was troubled by how many times this article used the word "deep". Also, what was the bit about the hack likely surviving a reinstall of the OS? Why in the world wouldn't it if it's a cpu bios firmware hack?
Deep thoughts with deep