CTS Labs Sent AMD and Other Companies a Research Package with Proof-of-Concept Code

[lewis007]

- should be "CTS Labs, the Intel-based IT security research company"

Could be Nvidia as well.

 

[Tumbler1987]

There's nothing to defend here when there are zero evidence and obviously the whitepaper barely makes any sense. You need a signed bios in one attack to run the malicious code.
It's like saying that MS is potentially distributing malware, because they sign their and their partners' drivers with a key and if that key is available, then you could sign your malware and spread it as MS software.
This has happened before and the key was published by mistake by microsoft... that's how you get those rights to produce and run signed s/w.

What has this become, the JIDF?
The guy said that Jew politics are shady, and reality shows that they claimed Palestine's land as theirs, they built a Wall to keep the natives away from their territory, they keep expanding their borders with various methods, including bombing and they kill on sight anyone who seems not to happy about those israeli soldiers who walk around with rifles.
You should not disagree with the jews, goy, that's antisemetic.


You have 0, that's zero, idea what you are talking about. Malware is many things, it's adware, trojan, virus, rootkit. Most malware doesn't run with root priviledges in many systems, it just needs a certain type of privileges to do its work. Most malware doesn't get planted magically on a PC, and it is usually a user's fault.
I could go on and on, but it's a lost cause, with people who have already shaped opinions and specific dislikes, not even mentioning the theoretical background.
Anyhoo, here's a simple example about the "rm -rf /" malware.
System: Loonix distro w/ systemd.
Systemd mounts Bios partitions in /dev/ and some versions mount it with write privileges for the root user...
If you "run rm -rf /" , you delete parts of your memory mapped bios. what does this mean? It means for for the motherboard to get bricked.
2 bios chips? you have a great chance that you will get it to POST in the next reboot.
1 bios chip? either you have to bring your soldering iron, or try an SPI programmer and there might be a chance for that motherboard.
How does this example align with the current situation?
You could for example take that file pointer from /dev/ and fill it with your "crafted bios"
then from the paper:
>Exploiting MASTERKEYrequires an attacker to be able to reflash the BIOS with a specially crafted BIOS update
nice, and how do you do that
>we suspect an attacker couldoccasionally still succeed in reflashing the BIOS
"suspect"... so you are not very sure.
let's go forth
>This could be done by first exploiting RYZENFALL or FALLOUT
nice, so I have to read Ryzenfall (that's a big claim there on the name) first
let's go to Ryzenfall's technique.
>Accessing the Secure Processor is done through a vendor supplied driver that is digitally signed
What?
So you tell me that asrock, asus, asmedia, american bios, can haz malware with their AMD supplied key?
so, you get the sign, you sign your code and this means that you can have either some microcode running on the cpu in ring -X, or on the bios itself.
How do you obtain the key? well someone has to provide it to you.
So there's no flaw there, a CPU, a peripheral, an embedded system _must_ run digitally signed firmware.
Where's the flaw? I have no idea? MS mistakenly had debug symbols on one of their supplied drivers, they lost a key, they invalidated it and its past.... we had good laughs.

But how come the almighty hack4z0rd W2zzard come up with "malware?
let's see the paper: a 20 page whitepaper that has zero facts, many mistakes, some ridiculous assumptions and some repetitive charts has 32 occurrences of the word "malware".
They are talking about the Arm Trustzone security "flaws", which ofc you have to exploit(if any) in order to gain access to the AMD PSP processor and there's hardly any mention of "Arm flaw".
There's some claim on twitter by one of the CTS guys, that ASMedia has an open windows on their firmware and there's no "ASMedia flaw" (they even say that asmedia's flaw exists because a few years ago asmedia lost a key :facepalm: )

conclusion:
there are some people out there, like w2zzard, who feel that it is their duty to bash some companies... and this has an impact on their sites. I found this thread because w2zzard wrote what he thinks is plausible, backlinks to amdflaws.com, then amdflaws.com backlinks here to say that this is a credible source(someone's opinion) so go read the "article on tpu or vice( uke: ) or other yellow sites.
If there's some PoC that does this, e.g. they get the key from a signed f/w with debug symbols still on the binary( doubt it coz of many embedded system reasons), then you just gain the ability to talk to the cpu. the Arm trustzone and the AMD PSP is well documented on some extend, therefore there's no "security by obscurity" as e.g. in the Intel ME where they didn't even said to the people that they ran on minix.
Unlike the PSP, ME has security flaws, that's why they found many parts of that system, that's how they found the OS it's running, the TCP/IP stack and so forth... that's how the documentation for the ME was written.
Arm trustzone IPs and f/w is available for purchase via Arm holdings. There's nothing to hide, the system has perfect documentation and there's a sh-tload of companies using it and debugging it. You set your own key and the system is secured.
That's a totally different approach and to be quite honest security by obscurity was a method they used in the 50s and 60s.

bonus:
the fail overflow team managed to get priviledges on the ps4 to run linux. how did they did it? there's a video on yt about this. The most interesting thing is that they had physical access to the board, a soldering iron, cables, programmers and a southbridge that was not made by AMD and had privileges to do IOMMU/DMA with the cpu.
They could easily claim "amdead" "ambankrupt" "$0 stock value" and what not, but they are professionals, first, and foremost they are people who have the background to do such exploits.
They knew it's not a Jaguar flaw when your southbridge rights on the memory of the system which is supposed to alter.

grats TPU, you gained, for another time, a few cheap Clicks.
clap();
wait(2000);
clap();
wait(2000);
clap();
return;

P.S.: I might be a Jew, you don't know. Disagreeing with me makes you an potential anti-semite. How does that make you feel? How are you going to sleep tonight?
P.S.: I wonder if there's an occurrence of code running so close to the metal and called malware. If my memory serves me right, this term is not used in bare metal situations and requires an OS and a security flaw to be called malware by researchers (those who submit papers, not those with a credit card and internet access to namecheap for a domain), but I am not sure. Food for thought anyways. See ya in several years again when another bubble hits the market.

Why so much vitriol? If you hate TPU so much, then you should read your tech news from other websites.

I should warn you though, get ready to be sorely disappointed, as the news on these flaws have been reported on almost every single reputable tech site, and even on non tech related news websites, so you'll have to deal with the fact that this is a newsworthy development, I would suggest avoiding reading about it if it rubs you in such a wrong way.

As far as W1zzard's credentials, he's been developing hardware level hacking tools since the days when Ati was still around, I remember using his OCing tools back when I had a Radeon 8500 in the early 2000s, so give the man some credibility, that gives him a fair advantage when formulating a well educated guess on the veracity of these alleged flaws.

Either way, this is a tech site, and TPU has every right to report on these very relevant news, they can't just turn a blind eye, or bury their heads in the sand while everyone else is reporting on it.

Idk why people are rushing to conclusions without even hearing from AMD, obviously this has caught the tech world by surprise, I strongly advise to wait for an official acknowledgement by AMD before formulating theories about the origins of the vulnerabilities, and the extent of damage we can expect from them if they're in fact real.

 

[EarthDog]

I think he takes exception to the delivery here, not the news itself.



Anyway, it may be a couple of days before we hear anything from AMD. As I understand things, third party validation took 4-5 days.

That said, I just noticed an update in the AT from this morning that may clarify one thing or another... or not.

Update 3/14 5:00am ET

Reported by Ars Technica, a second security firm has now spoken publicly about being contacted by CTS-Labs for verification of the vulnerabilities. Gadi Evron, CEO of Cymmetria, stated in a series of tweets that:

1. He knows CTS-Labs and vouches for their technical capabilities, but has no knowledge of their business model
2. All the vulnerabilites do not require physical access (a simple exe is all that is needed)
3. Fallout does not require a reflash of the BIOS
4. CTS-Labs believes that the public has a right to know if a vendor they are using makes them vulnerable, which is why no substantial lead time was given.

Quoted by Ars is David Kanter, founder of Real World Technologies and industry consultant, who verifies that even though these are secondary stage attacks, they can still be highly important. David states that while

"All the exploits require root access - if someone already has root access to your system, you're already compromised. This is like if someone broke into your home and they got to install video cameras to spy on you".​

Ars also quotes Dan Guido, who states that all that is needed to enable these exploits is the credentials of a single administrator:

"Once you have administrative rights, exploiting the bugs is unforunately not that complicated."​

​

 

[Xzibit]

I compare it to..

"Breaking an Entering" vs "Trespassing" your still not authorized to be there. Heck you can Trespass through a open door.

Well if i put a uniform and disguise myself as a cop (signed credentials) its okay right? What ever I do there after is there for a vulnerability.

Impersonating a public safety officer has very severe penalties.

Going to put a lot of strippers out of work

 

[FordGT90Concept]

Impersonating a public safety officer has very severe penalties.


How is Ryzen more adversely impacted than a Core i7 once root access is gained? In either case, I'd argue you already lost the war. Why do these 13 vulnerabilities matter at all? It's like comparing the aftermath of an ant invasion versus a earwig invasion. Full damage control mode response is more or less the same, no?

 

[R-T-B]

Still, when you have admin access, does it really matter at that point anymore?

When the malware can then survive a subsequent reinstall?

Yes.

The real vulnerability right there. What if I told you there is a vulnerability in the wild that allows anyone to do anything to a system no matter the OS. Its called the login/password.

That's been false since "hardware security" became a thing.

Or, it was supposed to be.

While everyone is speculating on this "revelation",
did we notice who did not get the package???

Rather telling don't you think?

Why didn't the Linux foundation receive it?
They're usually the first to come out with
corrective code. If we're doing a "public service"
why not give it to the parties most likely to
correct the problem first? Oh... it becomes
open source and everyone gets to see what
it is... or isn't... anyone ever try to get root
access on a Linux box lately?

Because linux doesn't run on the PSP?

 

[Prima.Vera]

All and all, I think this is good that all those vulnerabilities are disclosed now, or in the recent months, since all of those most likely were already known by the governments.

 

[R-T-B]

All and all, I think this is good that all those vulnerabilities are disclosed now, or in the recent months, since all of those most likely were already known by the governments.

I'm less conspiritorial than that. I know, I know, they've hoarded vulnerabilities in the past, but I'd like to believe complex ones like this to be beyond their needs or desires.

 

[lewis007]

I think he takes exception to the delivery here, not the news itself.



Anyway, it may be a couple of days before we hear anything from AMD. As I understand things, third party validation took 4-5 days.

That said, I just noticed an update in the AT from this morning that may clarify one thing or another... or not.

​

My question is why give Intel a 6 month lead in with their vulnerabilities and not AMD, Im no security expert but why give a company like CTS so much validity when they themselves state on their website that they make money from the corporations they are investigating, if that isn't a red flag then what is?

 

[jabbadap]

My question is why give Intel a 6 month lead in with their vulnerabilities and not AMD, Im no security expert but why give a company like CTS so much validity when they themselves state on their website that they make money from the corporations they are investigating, if that isn't a red flag then what is?

If you mean Spectre and Meltdown, those vulnerabilities was not just only in Intel's processors. Meltdown and Spectre were on IBM power 8/9 and some ARM manufacturers too. And Spectre was even on wider range of processors including AMD. And yes they were all given over the norm 90 days to respond on those vulnerabilities before making them public...

But yeah 24 hours is too short time to respond and makes all this very suspicious. Makes it feels like a stock manipulation scam disguised with real but not so severe vulnerabilities.

 

[EarthDog]

My question is why give Intel a 6 month lead in with their vulnerabilities and not AMD, Im no security expert but why give a company like CTS so much validity when they themselves state on their website that they make money from the corporations they are investigating, if that isn't a red flag then what is?

Cts answered that already. If you believe it or not is the issue...

 

[newtekie1]

You have 0, that's zero, idea what you are talking about. Malware is many things, it's adware, trojan, virus, rootkit. Most malware doesn't run with root priviledges in many systems, it just needs a certain type of privileges to do its work. Most malware doesn't get planted magically on a PC, and it is usually a user's fault.

I love posts like these. Ones where the person claims the other person has no idea what their talking about, then immediately goes on to make a statement that confirms what the other person said. It's so hilarious.

How is Ryzen more adversely impacted than a Core i7 once root access is gained? In either case, I'd argue you already lost the war. Why do these 13 vulnerabilities matter at all? It's like comparing the aftermath of an ant invasion versus a earwig invasion. Full damage control mode response is more or less the same, no?

It basically comes down to the management core built into the processor. When it was discovered that all Intel CPUs include an extra, basic, x86 CPU core that is initialized very early in the boot sequence that is responsible for managing all the other functions of the CPU, and it ran an open source LInux distro, it was immediately assumed that this was a security risk. If the Intel ME firmware could be replaced with something malicious, it would be very bad. However, Intel seems to have put in some very good checks to make sure this didn't happen.

Well, now we know that AMD has a similar extra CPU core that does management on their processors. However, they use an ARM Cortex based core. And it seems their checks to making sure the management engine isn't replaced with malicious content are not as hard to crack as Intel's. Allowing malicious code to much more easily be injected into the management engine. And once the malicious code is in the management engine, it is going to be extremely difficult to get out. These are low level functions of the CPU that normally have no real exposure to the OS, or even the BIOS.

That's all assuming that we believe CTS. Which I'm still hesitant to do. I believe they are doing this just to make a name for themselves.

 

[W1zzard]

memory mapped bios. what does this mean? It means for for the motherboard to get bricked.

you are talking about the bios shadow copy? that's not memory mapped but copied to system ram at bootup, and thus writes do not affect the flash chip contents

requires an attacker to be able to reflash the BIOS with a specially crafted BIOS update
nice, and how do you do that

talk to the flash chip directly, to erase and write nand pages directly, the same way the flashing software does it

>Accessing the Secure Processor is done through a vendor supplied driver that is digitally signed
What?
So you tell me that asrock, asus, asmedia, american bios, can haz malware with their AMD supplied key?
so, you get the sign, you sign your code and this means that you can have either some microcode running on the cpu in ring -X, or on the bios itself.
How do you obtain the key? well someone has to provide it to you.

I interpret "vendor supplied driver" as "the manufacturer of the chip has provided an (already signed) driver that's available to the public", that can be used for this purpose.
otherwise just make your own driver driver that does pci config space accesses through cf8, to find out mmio ranges and then write to these or map into user-space
the claim by cts is that the firmware hash checks on bootup can be bypassed

 

[Vya Domus]

Well, now we know that AMD has a similar extra CPU core that does management on their processors.

Was that a secret ?

 

[MarcusTaz]

This is clearly a scam by CTS-Labs to short the stock hoping it would tank and or get ready, "Opposition Research". Complete scam... They even put up a website amdflaws? hahahah How about a website called IntelSucksToo?

 

[newtekie1]

Was that a secret ?

I don't think most knew about it until very recently. And we definitely didn't know the inner workings of it, like that it was ARM Cortex based.

 

[W1zzard]

I don't think most knew about it until very recently. And we definitely didn't know the inner workings of it, like that it was ARM Cortex based.

I vaguely remember hearing about the Secure Processor being "ARM" during the first Zen briefings. Maybe it wasn't that widely communicated because wasn't really relevant until a few days ago.

 

[FordGT90Concept]

AMD was never quiet about Secure Processor being ARM. They actually say it's a Cortex-A5 in their white papers:
https://developer.amd.com/amd-secure-memory-encryption-sme-amd-secure-encrypted-virtualization-sev/

Of course we don't know what kind of operating system it is running. The vulnerabilities in Intel ME were exposed when it was discovered it was a branch of Linux.

Edit: Looks like Secure Processor might be open sourced. Maybe not all of it but:
https://github.com/AMDESE/AMDSEV


Edit: So yeah, AMD needs to push an update out for the Secure Processor to shut down the holes that were discovered. That sounds important, yeah, and another strike against the secure processor inside of the processor but it definitely doesn't warrant all the fanfare that was generated.

 

[Kaotik]

I vaguely remember hearing about the Secure Processor being "ARM" during the first Zen briefings. Maybe it wasn't that widely communicated because wasn't really relevant until a few days ago.

AMD has used and been open about ARM Trustzone based Secure Processor since 2013, it's not new in Zen.