Intel CPUs Are Vulnerable to New PortSmash Side-Channel Exploit

We expect that it is not unique to Intel platforms.
Click to expand...
Wait, Intel said this, right? Perhaps I am being a little dickish, but that comment from them feels a bit dickish itself, almost like they are pointing a finger at AMD and trying to push some of the heat onto them, a "it's them too and not just us!" sort of thing. Trying to throw AMD under the bus without throwing AMD under the bus. Not sure if I am making any sense LOL. :D
 
  • Like
Reactions: ltron
like this
ir0nw0lf said:
Wait, Intel said this, right? Perhaps I am being a little dickish, but that comment from them feels a bit dickish itself, almost like they are pointing a finger at AMD and trying to push some of the heat onto them, a "it's them too and not just us!" sort of thing. Trying to throw AMD under the bus without throwing AMD under the bus. Not sure if I am making any sense LOL. :D
Click to expand...

The researchers also said it. Don't know the technical details, but the attack is based on having a process run on a logical core, examining how it is being affected by a process on the other logical core on the same physical core, and then extrapolating data from that. Seems to me that such an attack can be used on any SMT-enabled processor, including ARM.
 
  • Like
Reactions: PaulP
like this
Tsumi said:
The researchers also said it. Don't know the technical details, but the attack is based on having a process run on a logical core, examining how it is being affected by a process on the other logical core on the same physical core, and then extrapolating data from that. Seems to me that such an attack can be used on any SMT-enabled processor, including ARM.
Click to expand...
Perhaps I am way overthinking what they said. Just felt like Intel was trying to throw the other CPU makers under the bus, since they [Intel] got so much crap [and continues to] over all the current issues.
 
I can't imagine disabling SMT really makes much of a difference in recent processors... regardless... shintel, lol.
 
velusip said:
I can't imagine disabling SMT really makes much of a difference in recent processors... regardless... shintel, lol.
Click to expand...

Depends on what you're doing. If you have 8 cores and what you're running doesn't max out 6 cores, it won't make a difference. If you're maxing out 8 cores, it can and does make a difference on the order of up to 20% depending on application.
 
Tsumi said:
The researchers also said it. Don't know the technical details, but the attack is based on having a process run on a logical core, examining how it is being affected by a process on the other logical core on the same physical core, and then extrapolating data from that. Seems to me that such an attack can be used on any SMT-enabled processor, including ARM.
Click to expand...

Woah, woah, woah. Don't bring ARM into this. ARM is magical perfection. Nothing can ever be wrong with ARM or make it have a flaw or weakness similar to x86.
 
Didn't have to even expand the headline to know that this was a cagey post/headline. Clickbait to the base at it's best. fans should read the article before posting.
 
Tsumi said:
If you read the article, it's an attack based on forcing a process to run on the different logical core of the same physical core.
Click to expand...

I know, aka shared resource timing attack. These have popped up repeatedly over the years, fyi. Its in every case a software issue.
 
ir0nw0lf said:
Wait, Intel said this, right? Perhaps I am being a little dickish, but that comment from them feels a bit dickish itself, almost like they are pointing a finger at AMD and trying to push some of the heat onto them, a "it's them too and not just us!" sort of thing. Trying to throw AMD under the bus without throwing AMD under the bus. Not sure if I am making any sense LOL. :D
Click to expand...

Not really dickish, but as a side channel timing attack using shared resources, there's a bit of history. There isn't anything special about any Intel architecture that makes it more or less vulnerable to this attack, any SMT is going to be vulnerable because by definition, SMT shares execution resources.
 
  • Like
Reactions: PaulP
like this
Jesus, the hits keep coming. Intel just can't catch a break.
 
GSDragoon said:
They probably just read the title.

I've been saying it work for years now. Security is inversely perportional to performance and usability. Security is why we can't have nice things.
Click to expand...

In one of my reports I wrote for college, I said that too much security would result in performance/speed similar to a "pregnant yak"... and I keep getting proved right over and over again.
 
sirmonkey1985 said:
as much as i enjoy watching intel get shit on for security flaws i feel this should of been tested on AMD processors as well before releasing this information..
Click to expand...

Ya this news seems a bit sketchy to focus on Intel when AMD can very well also be affected. Everything revolves around agendas these days.
 
Master_shake_ said:
what they don't know anyone with an amd system to test so they assume it is vulnerable too?

until verified it's not affected.
Click to expand...

From the guys that discovered this vulnerability: "A CPU featuring SMT (e.g. Hyper-Threading) is the only requirement".

So in principle any CPU with SMT is affected: Intel, IBM, ARM, SPARC, MIPS, AMD,...

They only tested two CPUs, but they are saying that all Intel CPUs are affected. It would be a bad security practice to assume that something isn't affected until proven the contrary. It is better to be safe and assume that something is affected until proven the contrary. If AMD CPUs are invulnerable, then AMD will make soon an official statement.
 
juanrga said:
From the guys that discovered this vulnerability: "A CPU featuring SMT (e.g. Hyper-Threading) is the only requirement".

So in principle any CPU with SMT is affected: Intel, IBM, ARM, SPARC, MIPS, AMD,...

They only tested two CPUs, but they are saying that all Intel CPUs are affected. It would be a bad security practice to assume that something isn't affected until proven the contrary. It is better to be safe and assume that something is affected until proven the contrary. If AMD CPUs are invulnerable, then AMD will make soon an official statement.
Click to expand...

Well ARM doesn't use SMT... so they are not effected.

MIPS and IBM Power could well be effected as both have SMT implementations. However I would bet good money neither is vulnerable, they have very different cache designs.

AMD its hard to say... indicators point to a big possible. On the other hand, it is also likely they didn't skip any userland privilege permission type test, that Intel seems to have skipped at so many important points. There is a handful of reasons Intels single threaded performance is higher, but a few of those reasons are terrible security design. I mean their in silicon cache system will honestly answer any program that asks before it even knows what level its running at. Deciding a userland program doesn't have permission after already answering them is just such a terrible design choice. It only makes sense in that the chip doesn't have to hit the shared bus for a microsecond to ask, for every speculation (as most will get tossed anyway) I get this attack is very different... still I will assume AMDs chip may well not be on the victim list until someone proves they are.

As you say though we'll know in the next day or so... I'm sure AMD and IBM will both be releasing statements of how this effects them one way of the other.
 
ChadD said:
Well ARM doesn't use SMT... so they are not effected.

MIPS and IBM Power could well be effected as both have SMT implementations. However I would bet good money neither is vulnerable, they have very different cache designs.
Click to expand...

ARM CPUs like ThunderX2 have SMT. IBM Z-series also uses SMT. We will have soon a list of affected/invulnerable chips.
 
I was wondering how far down the comments would I Have to read for the first one who at least read the summary if not the original article. I wasn't disappointed.
 
"Researchers say PortSmash impacts all CPUs that use a Simultaneous Multithreading (SMT) architecture"

"We leave as future work exploring the capabilities of PortSmash on other architectures featuring SMT, especially on AMD Ryzen systems,"

"AMD CPUs likely impacted"

:rolleyes: Either it does or it does not, which is it? Sounds to me like they were trying to lessen the impact on Intel by deflecting the problem, not taking the time to prove it. Therefore, they do not know and have no proof otherwise, until they test it.:vamp::wtf:

Edit: All these quotes are directly from the article.
 
after reading the article it sounds like if your a "average" desktop user then you have ALMOST zero chance of this affecting you in any way or shape whatsoever. the article also did not state if they did the attack through the internet or if it requires actual onsight access. i think if they are going to release stuff like this they need to cover every aspect of the hack, in detail. feels like a teaser......
no system is 100% foolproof, if it was then even the fastest system would seem like a 8086/88 speed and cost twice what they do now.
as for AMD, they informed intel a month ago, so they knew about it over a month ago yet they havent tested it out on a AMD system.....any credible researchers would cover all/any bases before releasing any data to to be reviewed.....yet they seem to not want to test a AMD system, why not?
i am sure that if AMD knew about this and wasn't affected they would have used it as a we are better tag line, and if it did affect them they also would have issued a patch.
 
Last edited:
ManofGod said:
Either they tested it or they did not, either it is proven or it is not. These are computers, not personal feelings.
Click to expand...

Yeah, but people are interpreting someone’s opinion that AMD maybe affected.

Like I said already, AMD is NOT affected because I own a 2700x.

Trololol
 
Cliff Notes:

Intel released a patch for latest processors
Side channel attack already requires PC to be compromised already. This just allows reading of privileged encrypted information.
It takes a long time for exploit to work.
 
01 Oct 2018: Notified Intel Security
26 Oct 2018: Notified openssl-security
26 Oct 2018: Notified CERT-FI
26 Oct 2018: Notified oss-security distros list
01 Nov 2018: Embargo expired
Click to expand...

when did they notify amd that their cpus were affected?

or did they pull it out of their assumption?
 
Dayaks said:
Yeah, but people are interpreting someone’s opinion that AMD maybe affected.

Like I said already, AMD is NOT affected because I own a 2700x.

Trololol
Click to expand...
the joke died after the first time.. repeating it didn't make it funnier.. but if your intent was to be an even worse version of Juan you're succeeded in joining him..
 
Dayaks said:
Yeah, but people are interpreting someone’s opinion that AMD maybe affected.

Like I said already, AMD is NOT affected because I own a 2700x.

Trololol
Click to expand...
Oh good, me just having the 1700 will protect my intels from these easy attacks.
 
Dayaks said:
Yeah, but people are interpreting someone’s opinion that AMD maybe affected.

Like I said already, AMD is NOT affected because I own a 2700x.

Trololol
Click to expand...
Someone being one of the researchers who discovered the vuln.

AMD is NOT affected Oh, You've tried the POC then?
Not tested just means not tested, nothing to do if they're affected or not. Which it sounds like they're testing now
(I thought your post before was /s but I guess not).
 
ltron said:
Then they shouldn't speculate before testing. The fact they have done this makes me feel it's more likely AMD aren't affected.
Click to expand...

Its not speculation, its simple logic. Timing attack that replies on shared execution resources = any processor that shares resources between contexts. They haven't explicitly verified it on anything but 2 models, but there is basically a 99.99999999999999999999999999999999999999999999999% chance that it will effect any CPU with SMT.
 
sirmonkey1985 said:
the joke died after the first time.. repeating it didn't make it funnier.. but if your intent was to be an even worse version of Juan you're succeeded in joining him..
Click to expand...

But it keeps working. I was drunk and felt left out. I never get to do the “it’s amazing because I own one.” Shit I even put trololol at the end.

It wasn’t a joke though.

And Intel is getting mauled by security flaws. Even if this one was real for AMD the mitigation would affect AMD less since they have higher core count at a price bracket, I would think?
 
I'd worry about this less than Spectre.

Of which I don't really worry about either.

These are flaws that require a very long time to discover any useful information because they depend on randomly happening upon useful data being processed while monitoring. Not exposing specific things to hackers at their will.

If you are running exploited software that long, you would have far more to worry about.

This is something for important people to worry about who have data important enough for this level of patient hacking, not for the vast majority of end users.

How successful this type of pattern analysis and inference of data will depend on how unique the delay caused by instructions / data are that's in the shared resource your thread is waiting on.

i'm sure if other cpu's are as easily vulnerable (or reliably vulnerable) intel will let everyone know. They're not going to want to be singled out if they can help it.
 
Oh, fantastic. Another round of fucking firmware updates for my servers at work. We JUST finished the spectre/meltdown ones last month.
 
NIZMOZ said:
AMD probably has the issue too. It says in the article AMD hasn't been tested yet.
Click to expand...

Nope, sounds like the article writer or researchers are just doing Intel damage control. Either AMD is affected or they are not, they cannot simply say they "may" be when they have done absolutely no testing that we are aware of. If they have done the testing, why have they not released the results?
 
I love that even when intel intros new hardware like the 9900K they still retain compatibility with exploits. :ROFLMAO:
 
To be honest I think anything can be hacked. Anything. Intel can try to design a secure processor, but someone out there will still find a way to exploit it, IMO.

My own opinion is that Intel needs to start developing an ARM CPU if they have not already done so. Same goes for AMD. For Apple to be nipping at their heels with a 5w cellphone CPU...it really spells trouble for me.

Actually I think Research in Motion/Blackberry would be a great acquisition for Intel at this point. And I am tempted to buy their stock at the right price. AFAIK no one has ever broken their encryption.
 
ManofGod said:
Nope, sounds like the article writer or researchers are just doing Intel damage control. Either AMD is affected or they are not, they cannot simply say they "may" be when they have done absolutely no testing that we are aware of. If they have done the testing, why have they not released the results?
Click to expand...

They understand the nature of the exploit, and they know the architecture of the AMD CPUs. It's likely that they know AMD is affected, but they haven't done the due diligence yet.
 
The proof code seems to rely heavily on the vpermd (permute) instruction. Neon does not have permute instructions so, much less likely.

It really comes down to how sloppy the register rename actions are. AMDs saving grace may be that they do not execute permute instructions nearly as fast. LOL.

Edit: I reread the code and the above is probably not exactly what is going on. vpermd is the instruction they are smashing but the registers are interleaved in a way that makes no sense to me. My armchair is armchair.
 
Last edited:
You must log in or register to reply here.
Back
Top