Hacker News new | past | comments | ask | show | jobs | submit login
Intel platforms from 2008 onwards have a remotely exploitable security hole (semiaccurate.com)
506 points by theSoenke on May 1, 2017 | hide | past | web | favorite | 180 comments



The short version is that every Intel platform with AMT, ISM, and SBT from Nehalem in 2008 to Kaby Lake in 2017 has a remotely exploitable security hole in the ME (Management Engine) not CPU firmware.

We knew this would happen. We knew that the Management Engine was a backdoor, and we knew it was only a matter of time before someone would figure out how to exploit it. This is exactly the reason why Libreboot exists (https://libreboot.org/faq.html#intel). And now, far from being the tinfoil hat distro that is often portrayed, it will become a bare necessity.


This is also what the management engine cleaner project is for:

https://github.com/corna/me_cleaner


https://github.com/corna/me_cleaner/wiki/How-to-apply-me_cle...

The procedure seems far from trivial and requires special hardware(?). Is there a guide or some resources I could follow as a person with no hardware/low-level technical knowledge?


There is no good solution for Intel chips.

You could sidestep the whole issue by buying a C201 chromebook (quad-core ARM) and putting Linux on it.


You're not kidding!

> Internal flashing with OEM firmware

> --------------------------------------------

> TODO


Let's hope one of the other CPU manufacturers (e.g. AMD) starts supporting LibreBoot and allows to officially disable the ME-equivalent hardware feature, so that Intel get's forced by market-pressur to follow.

Intel needs more competition - thanks to AMD latest new 8-core CPU Intel got forced to release a new CPU the had in their basement for years - suddently it's possible for them to release i7 notebook CPUs with more then two cores!! Even back in 2010 it would have been viable to produce 4 core notebook CPUs - but the went away because the had no competition.


That was the top request in their March AMA:

https://www.reddit.com/r/Amd/comments/5x4hxu/we_are_amd_crea...

I wouldn't hold my breath, though.


The sad thing with that is that

- releasing the source doesn't tell you what's on the chip.

- PSP is kind of "Ring ∞", so there would be no good outcome from providing general-purpose access to it. So, the keys will never be released.

- it's thusly not possible to map the signed (encrypted) firmware to the source.

- even if the source had a clearly documented "master off" in it, you can never know if the firmware's copy reads "master-except-if-A-and-B-say-C off" :(


What are you on about? I had a 4-core i7 in my laptop back in 2013, an i7-3920XM IIRC:

https://ark.intel.com/products/64887/Intel-Core-i7-3920XM-Pr...


> suddently it's possible for them to release i7 notebook CPUs with more then two cores

I'm not sure what you mean by this. My Dell XPS 15 has a i7-6700HQ which is quad core, and it's not like I just bought the thing.


> release i7 notebook CPUs with more then two cores

U-series i7s have two cores. HQ-series i7s have four cores. Both are mobile CPUs. Remember though that more cores generally means more power consumption which generally means less wallclock time on battery power.


Intel's U-8XXX CPUs are rumored to offer 4 cores with a variable TDP from 18-45W this autumn.

It's my tinfoil hat theory why MS waits for an earnest update on their highend Surfacebook. A high-end quad-core Surfacebook with a 10-series GPU and 32GB LPDDR with real Thunderbolt 3 Ports would make for a 13" dreammachine...


first mobile quad cores were sandy bridge released january 2011


? Nehalem had mobile quad cores. I'm using one right now.


Specifically the Clarksfield processors from 2009: https://en.wikipedia.org/wiki/Clarksfield_(microprocessor)

Predating the i7 entirely, there were also quad core laptops using Core 2 Quad CPUs (Penryn QC) in 2008.


Indeed (have some of those too), but I assumed we'd artificially limited ourselves to talking about i7. I'm not sure why I assumed that, because you're right of course.


I'm having fun, I finally have an excuse to dust off my Libreboot X200 (refurbished and modded Thinkpad with Libreboot firmware).

However, I strongly disrecommend buying from Leah Rowe unless you enjoy waiting months for payment confirmation and delivery. The worst webshop experience I've ever had.

I recommend you build/flash your own, contract it out or look for a different vendor.


Has anybody tried the X200 builds from Libiquity?

https://shop.libiquity.com/product/taurinus-x200


I don't own anything by Libiquity but I'd be happy to ask answer any questions about my Libreboot/Minifree X400, which is based off the Lenovo Thinkpad T400, including diagnostics and/or benchmarks if I can get them set up in less than 30 minutes.


If the verilog to the chip isn't open, you can't trust it. Stallman is dangerously wrong on this point.


Somewhere you have to externalize trust. What use is the open HDL code for a chip if you cannot be sure someone down in the manufacturing chain hasn't... modified it?

Certainly this kind of attack is not your average script kiddy but nation-level instead, but I wouldn't put it past the NSA to pull this off.


Correct, you do need to externalize trust somewhere, but the Richard Stallman level of "chips are ok but firmware is not" is not the correct place for it.


If only we could checksum the commercial hardware and compare it to a reference implementation checksum.


> For obvious reasons we couldn’t publish what we found

It's not obvious to me why anyone not under an NSL or NDA would sit on this vulnerability for 5 years and wait until it's actively being exploited in the wild before public disclosure.

It's extremely negligent to global security for SemiAccurate to not immediately publicly disclose the vulnerability 5 years ago after Intel refused to fix it. Of course this is ignoring the root of the problem, which is that the US government has deeply compromised Intel since the very first security management interfaces were added to Intel chips in the early 90s.

The real solution to the root issue is legislation that forces security disclose timelines of 90 days or less for government-found vulnerabilities, and prevents the stockpiling of vulnerability exploit kits.


That seems strange. Since it's a security hole you can exploit on your very own Intel computer, there's no issue about "hacking" into someone else's system. Researching this is legally safe. There should have been a Defcon talk and a CERT advisory years ago.


It gets more confusing because Intel is crediting Maksim Malyutin from Embedi: https://security-center.intel.com/advisory.aspx?intelid=INTE...

Intel would like to thank Maksim Malyutin from Embedi for reporting this issue and working with us on coordinated disclosure.


I interpret that as SA got wind something like this was going down, guessed some of the details and possibly forced Intel to disclose but they didn't actually find anything themselves nor do they have the details. Which explains why Intel credits someone else and they overplayed their hand by claiming that either ME or VPro are breached when it really is AMT. (Bad enough...)


This was my thought too as I read it. If they didn't feel they could handle the disclosure, Google Project Zero could have been a good recipient to report to Intel.


One wonders if they knew of this particular vulnerability 5 years ago or they just knew that there must be vulnerabilities lurking in the ME somewhere.


Is there a better source for this than SemiAccurate? The article doesn't really have much beyond self-aggrandizement and "we can't tell you any details, but you're screwed". For something that could be anything from "Charlie Demerjian heard a rumor about a ME patch and wanted some pageviews" to the actual security apocalypse, I'd like credible sources.



From the Intel advisory:

> There is an escalation of privilege vulnerability in Intel® Active Management Technology (AMT), Intel® Standard Manageability (ISM), and Intel® Small Business Technology versions firmware versions 6.x, 7.x, 8.x 9.x, 10.x, 11.0, 11.5, and 11.6 that can allow an unprivileged attacker to gain control of the manageability features provided by these products. This vulnerability does not exist on Intel-based consumer PCs.

If in doubt, you can check your CPUs here:

https://ark.intel.com/#@Processors


I'm glad that credible sources are now available. It's unfortunate that it took so long to confirm, but congrats to SemiAccurate on a massive scoop.

Edit: On the previously-mentioned scale, this sounds like a solid 8 or 9 out of 10.


Credibility issues of the author/website aside, I actually hope this is true, and I hope it's catastrophic for Intel.

Maybe then we'll finally see hardware companies taking security seriously.


I'm worried that it's true and it's not catastrophic for Intel. Aka show to the world that you can get away with BS like this.


The fact that people can stay behind platforms, companies, and technologies that are proven to be so inherently insecure that they can never be trusted just boggles my mind.

Adobe Flash has a new zero-day every week, but we were saddled with it for years past when it should have been retired because some people didn't want HTML5 to have feature-parity with Flash.

Java has a new zero-day every week but we're stuck with it because enterprises are afraid of trying something new.

Windows was wide open to attacks for years, but they got away with it by saying "yeah but Apple is so expensive" and people still parrot that. They said "yeah but Linux is stolen technology/doesn't work right" and people still parrot that.

Android has a new malware/exploit warning every week, the majority of the phones never see security updates, and are running outdated software the minute they're shipped to stores but people say "yeah but Apple is so expensive/locked down" or "Windows Phone doesn't have any apps".

I have friends who lost their credit card numbers at Home Depot but refuse to shop at Lowes because they don't like the NASCAR driver that Lowes sponsors.

People get so caught up in brand loyalty that they're willing to defend "their" company like it's a family member. Even among the tech community, security means nothing. We still use Android phones to get root access, we still use Windows to save some money on our laptops, we still program in PHP because it pays the bills.

Nothing will ever be catastrophic enough. Anyone can get away with it just by creating an "us vs them" mentality with their customers.


Java has a new zero-day every week

No it doesn't. The last one was in 2015. Before that I think there was a two year gap to the prior one. Zero days in Java are actually very rare these days.

That doesn't mean bugs are rare - like any large piece of software Java gets regular security patches, but those are flaws found by the developers themselves rather than attackers, so they aren't zero days.


I think it's implied that "a new X every week" is always going to be hyperbole. I'm intentionally overstating the point so someone just like you could hop in and prove it better than I ever could.

Remember in 2012 when Apple stopped shipping Java with their browser because it was so insecure?


Upgrading Java 5 minutes, 0 $

Upgrading your CPU 2hs, 300$

Having no secure CPU to upgrade to: priceless


>> People get so caught up in brand loyalty that they're willing to defend "their" company like it's a family member.

Long ago I read something about that. The psych came down to the (false) idea that changing brand would confirm that you were wrong. The example was that even if Ford made better cars back in the day so you're a diehard Ford owner, if they quality demonstrably falls behind and Chevy is demonstrably awesome today you still won't change! And that's a case where your prior decision was actually right. So people have these weird internal notions that 1) companies value doesn't change over time, 2) their value doesn't change in light of new evidence, and 3) My own value is somehow tied to making a "correct" decision in spite of cognitive errors #1 and #2.

People are stubborn, and that's being kind about it.


>Java has a new zero-day every week but we're stuck with it

Well, Java applets did die. What more do you want? The Java sandbox is only used by extremely legacy software at this point, so it doesn't matter if it has holes in it. Actually, the more holes the better, so we can get rid of the last holdouts.


Java is the most widely-used programming language in the world. Applets are an insignificantly tiny drop in the bucket of what Java is used for.


Java is consistently listed in the top three (and often #1) languages in current use.


IME is likely not a case of Intel "not taking security seriously". It's almost certainly a case of doing what FiveEyes demanded of them.


You're probably right.

I still hope it's true, and that it's catastrophic for Intel. No change can happen otherwise.

If Intel aren't fighting against 5eyes then they aren't taking security seriously.


this was my first thought as well, but surely there would have been some hint of it in snowden docs or the recent wikileaks cia malware docs?


I'm not familiar with the author. Can you elaborate on the credibility issues?


Charlie Demerjian is a massive hater. That doesn't mean he's wrong, but everything he writes about Intel or Nvidia has a negative slant.


There's eventually going to be one when it is officially published by Intel, but that seems to be months away right now.


No, that's not how sources work. You don't get to use your assumption that the article is accurate to assert that it will eventually be proven accurate by other sources. That's circular reasoning.


If the article's claims are true, all sources (e.g. OEMs with access to a fix) should be under NDA, https://twitter.com/cdemerjian/status/859096565033693185


...and if the article's claims aren't true, there wouldn't be any sources to confirm the claims at all. The evidence we've been presented with so far (no sources) is consistent with both possibilities. When you make a claim as big as SemiAccurate did, it's on you to provide sources to back it up. If you can't present any kind of proof, you don't have a story, you have a rumor.


The article claimed:

> That is the end of June for non-Intelspeak people, they will officially issue this guidance then along with OEM disclosures.

We'll know in two months whether the above claim is true or false.


My prediction: At the end of June, Intel announces a fix for a minor non-RCE bug in the LAN code of Intel ME. SemiAccurate proudly and inaccurately announces that it confirms their previous reporting and adds it to the list of things to mention every time they write an article about Intel. There is no follow-up Hacker News thread with 100+ comments, so most of the people who posted here continue thinking that there was a major RCE in Intel ME that we just haven't heard about because it was covered up.

Edit: Already proven wrong! We're headed for interesting times.


https://security-center.intel.com/advisory.aspx?intelid=INTE...

Looks like an almost full confirmation. They're saying consumer hardware is unaffected, but everything else matches.


Yeah, I blew that prediction pretty badly. Congrats to SemiAccurate on the huge scoop!


Then it's a rumor. One you probably want to keep your eye on. Which was the whole point of the article anyway.


Devil's Advocate here, so you are assuming there is a perfectly secure software implementation in this world, and only Intel has it for their Management Engine? I get your point, SemiAccurate may or may not have an exploit, but I think it goes without saying there is a security hole somewhere in the ME, it just is not publicly known at this point.


If Intel released a firmware update, then anyone can compare this update to a previous version and see what has changed.


That's harder in practice than you make it sound. Firmware updates for Intel ME are handled through OEMs, it's not a file that Intel publishes that an interested person can go to their website and download. The article claims that such a patch has been released to OEMs but is being kept under wraps, which might make it hard to determine when it actually ships in a downstream update. Even if you have a file that you know contains the binary blob of updated firmware, reverse-engineering it to determine what it does differently compared to the previous version is very much non-trivial.


CPU firmware patches have also been released through Microsoft update, and Windows computers may well be patched on the fly in this way. I suppose it would be possible to download the update individually and examine its contents, but, as you point out, it would be extremely difficult to work out what it was doing.


Is there a better source for this comment than "I don't like Charlie Demerjian"?



That is not a source that confirms SemiAccurate's claims, it's a nearly decade-old paper (~2008-2009) describing an attack against a completely different security feature. The "System Management Mode" described and attacked in the paper is unrelated to Intel ME.


What is the publication date of this?


Initial confirmation below, I haven't read it through in full yet. I believe technical details are being delayed;

https://security-center.intel.com/advisory.aspx?intelid=INTE...


>The details of our new SMM attacks will be made available once Intel patches its firmware, most likely we will present them at the Black Hat USA conference in summer 2009. We will also make the code of our TXT exploit available.


>>every Intel platform with AMT, ISM, and SBT from Nehalem in 2008 to Kaby Lake in 2017 has a remotely exploitable security hole in the ME (Management Engine) not CPU firmware.

>>there is literally no Intel box made in the last 9+ years that isn’t at risk

>>SemiAccurate has been begging Intel to fix this issue for literally years

Am I the only one who is so cynical to think it must have been deliberate? Intel dragging their feet for YEARS -- what could justify such a delay? The paranoid side of me asks "Were they waiting to patch this hole, until they found a different one that could be utilized?" Which begs the next quesion: Where is the NSA in all of this? It's the sort of thing that would be mighty handy to a group wishing to snoop on everyone and everything?

Last question: Why would anyone trust the encrypted management engine after this? (Why would anyone trust it before?)

>> What about embedded devices that are increasingly PC based? Digital signage perhaps? Industrial controls. HVAC. Security systems. Flight controls. Air traffic controls. Medical devices.

What, indeed? Is this the method used to interfere with Iran's nuclear program centrifuges?


Extending the attack surface in the name of alleged convenience seems more a plan to enable hacking than a reasonable design and marketing approach for microprocessors. IoT fanboys with an urge to make home appliances remotely exploitable might be in good faith, but Intel is smarter than them.


Believe incompetence before malice, and I'd stick economic incentives somewhere in the middle.

The discussion probably went something like:

Person 1: "Should we issue a recall and disable a feature which bought us a several billion dollar customer?"

Person 2: ...


In this day and age the choice between malice and incompetence seems to fall more on the malice side.


reminds me of CIA's Simple Sabotage Field Manual[0]

https://www.cia.gov/news-information/featured-story-archive/...


As a sysadmin at a Windows shop, I don't know what to make of this. Has Intel commented on this, yet? Any OEM?

Joanna Rutkowska, who is a renowned security researcher, warned of something like this happening sooner or later[1], so I don't think I can afford to just ignore this.

But without something more specific to act on, there is nothing I can do, except wait firmware updates to be released by various vendors. If that happens.

And what if Intel does make a statement that essentially says, "This is all total BS"? I wouldn't know whether to believe them or not.

The only scenario where I could have any degree of certainty would be if Intel came out and said, "Yeah there's an exploitable security hole in ME, here's a patch to disable it".

[1] http://blog.invisiblethings.org/papers/2015/x86_harmful.pdf


As pointed out by another commenter, Intel has released the advisary:

https://security-center.intel.com/advisory.aspx?intelid=INTE...

It confirms much of the SemiAccurate report, but also includes this:

"This vulnerability does not exist on Intel-based consumer PCs."

Which seems to differ from what SemiAccurate was saying. I'm not sure if it's SemiAccurate being... er... not completely accurate :D, or if it's Intel trying to downplay things.

I guess we'll find out more over the next few days/weeks.


Looking at the Intel link, they take you down a path to see if you have vPro. That's on some i5s and i7s. So they are defining "consumer" roughly as "purchased at best buy or similar". There are certainly desktops in people's homes that have vPro. Even some of the higher end NUCs have it.


Easier path: https://ark.intel.com/#@Processors

When I've purchased VirtualBox hosts, I've deliberately avoided stuff with vPro.


I assume this is a bug in firmware version that's only used on enterprise class hardware, but that they assumed the bug also exists on consumer grade hardware on the basis of the circuits being mostly the same (and thus assuming the firmware is shared).


Believe it in proportion to the supporting evidence presented. At the moment, that's nothing except an appeal to the widespread belief that an Intel ME security flaw is inevitable.


Update, Intel has now confirmed that there's an issue: https://security-center.intel.com/advisory.aspx?intelid=INTE...

They claim that it doesn't affect consumer CPUs, but that leaves a ton vulnerable. It's pants-shitting time.


This thread (and the link for it) have some decent information, for those looking for it. It had a cross-link here, figured I might as well link back to it: https://news.ycombinator.com/item?id=14242508



Zero details and zero cross references, zero mentions on Google and zero mentions in any security list I'm on. Charlie blowing nonsensical steam yet again?


The article implies that they have been privately trying to get Intel to fix it, so there is no reason it would have been mentioned publicly anywhere.

Now a patch is coming out but Intel is still trying to keep it quiet, so he's trying to warn people disable AMT and be ready to apply patches ASAP.

Presumably he didn't even want to disclose the existence of the vulnerability publicly until there was some sort of fix, and he still won't want to disclose details before the fix is released.

Of course, you can doubt the veracity of this story, but I'm just pointing out that there would be no reason to expect details, cross references, or mentions on Google or security lists yet if it is true.


If Charlie was a security researcher and SemiAccurate was a well-regarded security firm, I would not expect details or cross-references or mentions on security lists. Charlie is not a security researcher, he's a journalist, and SemiAccurate is the tech equivalent of a supermarket tabloid. He is not a credible primary source for anything security-related, particularly given SemiAccurate's reputation for publishing rumors as facts.

None of that means he's necessarily wrong, just that you should be very careful about believing his claims without supporting evidence. A lot of people here on HN have thought that a remote ME exploit was only a matter of time, so an article claiming to validate that belief will not get as much skepticism as it should.



It does seem suspicious to me that this hugely critical flaw deep in the firmware stack has been discovered by the writing staff of a tech news website rather than an infosec research team...


The article sort of reads like he has thought (not known) there was an issue for a long time.

Then, he saw that Intel released a patch related to the management engine, and took that as confirmation? Maybe he has access to the release notes via a source at an OEM?


He got the affected version numbers exactly right, but according to Intel he got the affected hardware wrong (consumer hardware unaffected).

That tells me he got the information from an unmentioned source. If he had the details himself he would be able to confirm what hardware it is present on by testing it.

The explanation for that error code be that the source have been vague about it or not tested it on a lot of hardware, or isn't even a firsthand source, or that the journalist misunderstood it.

https://security-center.intel.com/advisory.aspx?intelid=INTE...


Ah. To be fair, that intel.com link is confusing, because it sends you off to see if you have "Intel® vPro", which is certainly on consumer hardware, like various i5 and i7 systems. Which does not jive with the earlier line "This vulnerability does not exist on Intel-based consumer PCs". It sort of depends on your definition of consumer hardware.

Overall, though, it does seem to validate the sequence was something like "he had a suspicion" then "intel released an update".


Is there any way to avoid the patch and reverse engineer it to "root" ME and cripple it?

I'd guess this would be a lot of interest for "hacker" news; i want to sign my own damn firmware.


Yes, but we've only figured out how to do that for old hardware.

Check out https://libreboot.org/docs/hardware/gm45_remove_me.html


Yes. It is his uncontrollable urge for getting thousands of corporate IT admins to disable the Management Engine, at it again.




I think it is high time for companies who make hardware be financially fined for lapses like this. In this particular case, the manufacturer was warned and did nothing for years.

This is negligence especially considering these chips control critical devices that can cause damage or even loss of life if they are successfully exploited.

Can you imagine if car maker didn't fix a hardware defect they knew for years. Oh wait...


What is the motivation behind Management Engine?

From the perspective of an everyday user these things came out of nowhere to evolve into this para-computer running along side me that I cannot see and have no control of. It is on literally ALL hardware

Why is it that any attempts to disable it knock your whole computer out?

And this is the world of technology that we want? I'm so sick of technology companies appearing to work for their customers but secretly working against them.


The functionality ME attempts to provide is lights out a.k.a. out-of-band management (like IPMI) to the desktop.

If, for example, an admin needed to add a dual-boot-to-Ubuntu option to every PC on a floor, he could, through ME, remotely reboot (force power reset if necessary) or power on every machine, have the machines boot to a (remote) OS install disk, run the install, and reboot.

ME allows one to do almost anything remotely to a PC, regardless of what the main processor is doing. That is both useful and frightening.


Fine, but putting it on all hardware?

How many corporate IT environments buy off-the-shelf motherboards and CPUs from the same channels as consumers? OEMs get an entirely different set of parts and enterprise sales works in completely different channels. If there is such a clean separation between corporate and consumer markets then why is this hardware on everything, and why does it need to pull power on the machine if it's disabled?


It isn't on all hardware. Intel has two ME firmwares, a small one for consumer systems, and a big one for corporate/enterprise systems. The small one does not (or at least, should not; is not supposed to) include the remote management features.

In other words, the separation that you describe exists.

Systems with the full firmware sport things such as the vPro branding, and only certain combinations of CPU and chipset support it.


AFAIK the consumer version still kills the system if it's disabled?


I'd be careful with assumptions on what "consumer hardware" means. There are desktops, NUC units, etc, that shipped with i5 and i7 chips that had vPro.


Even with the CPU, you also need the right chipset and the right firmware to actually light this stuff up. While especially in the laptop sector there are consumer devices that include this, it's far from universal.


Can't all that be done from the main OS? Repartition, modify the boot stuff, reboot from an image in a new partition, etc... Why did they need to add another processor with closed source and all the potential security issues?


You can't change the boot media or turn on a turned-off machine via the OS. The whole point is to get underneath it, so you can even do initial OS install with it.


It might not be trivial, but you can do this w/o the ME. My understanding is that most ethernet cards support a "Wake-on-LAN" feature to turn off machines on, and from there you can trigger the machine to reboot and then netboot (by writing to its boot config to instruct whatever boots it that it should take that action).

Even if you assert that the ME is absolutely necessary for such a use-case, I don't have that use case, it isn't work the risk for me, and I should be able to disable the ME because I, as the owner of the machine, want to. (Or really, otherwise interact with it and use it for creative use-cases.)


Just get a computer that doesn't have vPro.


That disqualifies a lot of otherwise really good hardware. My current Thinkpad, for example, and all current MBPs, I believe. Some manufacturers also aren't very clear about the exact hardware in their machines, either. (For example, Apple doesn't list the exact CPU on their tech specs page, only the somewhat vague "2.4GHz dual-core Intel Core i7, Turbo Boost up to 3.4GHz, with 4MB shared L3 cache". That might be unambiguous enough to map back to an actual piece of hardware, but it's still a considerable amount of work to do so.)


I was curious how much work it would be, so headed to ark.intel.com.

The filter for "Max Turbo Frequency" seems to be broken, BUT searching for "Cache: '4MB L3 SmartCache'" (which I assume means shared!) finds quite a number of results: http://ark.intel.com/Search/FeatureFilter?productType=proces...

The list is column-sortable by Max Turbo Frequency, and there are just 4 results in the 3.40GHz range:

- http://ark.intel.com/products/91169/Intel-Core-i7-6660U-Proc... - the only result with a Processor Base Frequency of 2.4GHz

- http://ark.intel.com/products/91497/Intel-Core-i7-6650U-Proc...

- http://ark.intel.com/products/88192/Intel-Core-i7-6600U-Proc...

- http://ark.intel.com/products/52231/Intel-Core-i7-2620M-Proc...


I think the problem is not that this technology exists but rather that the operation of this engine is not transparent, the user cannot examine or disable the software in this engine, cannot write his own software.


IME should exist on an external TPM chip so it's only for those that want it, like enterprises.

I really don't understand why the would just shove it into every chipset out there. I understand it needs to get its claws all over the system, but the core should be external and optional.


Because a separate chip costs more.


"There is anything to worry about if you have nothing to hide" /s



   Security is a cost center and most OEMs run on margins too thin 
   to bother with security patches even if they cared. Most simply don’t care.
I think that sums up pretty well why downstream vendors are treating security casually. So the billion dollar question is, how do we fix this, as a tech community?


This is an unpopular position, but approaches like BrickerBot are likely to be effective.


OEMs are not involved at all with ME afaik, it's exculusively controlled by Intel.


OEMs have to ship ME firmware updates; Intel has no way to get them to you directly.


Can't they install an update remotely via this vulnerability? :p


No joke, this would be the best thing for everyone. Especially if we find a way to do it ourselves rather than wait for a vendor to.

I've been thinking for years about writing a virus that patches the vulnerability it used to spread as it goes.


Open architectures are a solution, even if there is no single common solution. Diversity is something we have been missing since windows became popular, and although security through obscurity is not a strategy, diversity certainly serves well at limiting the scope of damage possible for a single attack.


I'm not sure the tech community is able to fix this, short of the brickerbot mentioned by another poster. Frankly, I think this situation will only resolve if and only if there are dire financial consequences to OEMs that pay lip service to security.


Great news that this finally came to light.

After learning about remote management capabilities I've always suspected it had holes. Large attack surface, any exploit would have a high value, and closed source.

Perhaps one day we'll be able to buy CPU's without this "feature". I'm betting AMD and ARM are in the same boat.


> After learning about remote management capabilities I've always suspected it had holes. Large attack surface, any exploit would have a high value, and closed source.

Even after reading this, I'm still not convinced it does have holes. It's so high value (pervasive, incredibly powerful, and old) that if it were possible a bad actor would have used it. The spectrum of possibilities is small:

    1. The hole does not exist, but SemiAccurate thinks it does.
    2. It exists, but only SA has discovered it.
    3. SA discovered it along with a few bad actors, who are using it surreptitiously and haven't been caught.
    4. It's being used all over the place, it's a widely acknowledged security disaster.
We're not in state 4. The article suggests we're in 2 or 3. 2 seems unlikely - SA does not have special abilities that transcend those of other security research firms. 3 seems especially unlikely: with this much power available, and with the hole being patchable, could they resist using it? Which leaves option 1.


SemiAccurate isn't a security research firm, it's a tech news blog. There's basically no chance that they've discovered anything. If there's an exploit, they would've had to have heard about it from either a source inside Intel or an actual security researcher of some kind.


There are many high profile targets where "if it were possible a bad actor would have used it" has been proven false. See recent publicity about vulnerabilities in printers, antivirus products, etc.


why would 3 be unlikely ? The snowden leaks indicate that 3 letter agencies will use high value 0-days sparingly, to ensure they remain 0-days.

Also, there is a 5th (more likely) possibility: SA didn't find anything, but undiscovered holes do exist.


https://security-center.intel.com/advisory.aspx?intelid=INTE...

5. Somebody else discovered it and told SA. No idea why them rather than telling anybody else.


They are. AMD TrustZone runs an ARM core alongside your computer. I've also heard a lot of ARM SoC platforms have something similar.


The key difference is that this doesn't affect every ARM processor.


"It is this last point that has been causing some political unrest in the US, and the rest of the Western world. As you undoubtedly know, China is very nearly the sole producer of all electronic goods. It would be very, very easy for the Chinese government to slip a hardware backdoor into the firmware of every iPad, smartphone, PC, and wireless router." 2012 https://www.extremetech.com/computing/133773-rakshasa-the-ha...

Made in China, designed in the USA. Everyone wants their own backdoor.


Patching is going to be a nightmare considering that many OEMs drop support for a motherboard after 3 years. There will be unpatched systems floating around for a very, very long time.


I've got a Lenovo T530 and a Lenovo T450s. I wonder if they've released a firmware update yet...?

I can't say I'm surprised, but I am surprised at the fact that finally, after all these years, someone finally got down to patching some vulnerabilities in this area.

props to whomever forced Intel's hand.


One nice feature of (some) Thinkpads is that the AMT and ME can be "permanently disabled" through the BIOS, presumably by blowing a fuse or similar. Check if yours has this capability.

Otherwise check for updates at http://pcsupport.lenovo.com.


Hopefully, someone can speak further to whether this is a real mitigation and what "permanently" and "disabled" really mean, in this specific context.

I'm don't mean to sound oppositional. I appreciate this being mentioned.

I'm just not willing to trust it without knowing in detail that and how it works.


Can anyone add any details? The article is very very vague. Doesn't this work thru the Ethernet port in the chipset silicon?

So if you're running a desktop that has a physical Ethernet card in it, and the Intel Ethernet isn't connected, are you OK?

And if you're running on a laptop that uses Intel's Ethernet, (and most of them do?) then are you vulnerable?


Worrying about the ME and my dislike of secure boot is what has kept me from upgrading beyond the Core 2 Duo with BIOS. It's starting to feel slow now, but I still don't feel I can upgrade unless there is at least a way to disable the ME. So far, there don't seem to be any reliable methods of doing so.


Even without any newly discovered backdoor. The Intel ME was always a fuing security issue. A BACKDOOR. It is completely naive to think the NSA can't use the ME to get access to anything, but hey it needs another Snowden for people to listen again.


Intel ME always reminding me of the saying, "Absolute power corrupts absolutely."



"The Intel ME subsystem can take over your machine, can't be audited" - OP's discussion's title

> do the first three steps of thinking for them. Make it really easy for the other person to say yes or no

source: http://firstround.com/review/how-to-become-insanely-well-con... | https://news.ycombinator.com/item?id=14195664

I've agreed to include a bit of detail when spamming all my friends links via e-mail going forward.


For those who cannot switch to Libreboot, https://github.com/corna/me_cleaner may be a solution to this issue.


What is the management engine, and how does one access it remotely?


it's a closed-source binary blob on intel chipsets with unfettered access to the CPU. it is also (often) directly connected to the RJ45 port.

here's a good overview of the risk: http://hackaday.com/2016/11/28/neutralizing-intels-managemen...


So if you don't use the RJ45 port on the motherboard but instead use an RJ45 port on an expansion card instead you're safe?


Partially. Expansion cards use PCI-E which has DMA capability, so a bug/backdoor in their firmware can very well be used to attack a system.

But I believe newer systems with MMUs acting as "firewalls" for DMA are safe from this vector.


there's also the concern of physical attacks, via the motherboard's RJ45 or USB.


At least USB doesn't have device-initiated DMA, but USB descriptor parsing bugs have in the past led to exploits (I remember the PlayStation jailbreak).


A good argument to epoxy those ports shut, if you're really worried about that.


Does this affect an Apple MacBook?


Assuming that what the author says is true and there is a local exploit for non-enterprise versions of the Intel ME, then yes.


If it does, can EFI patch it out?


It'll be interesting to see how Intel deals with it.

Looking at the recent Atom failures (with vendors told in no uncertain terms to present publicly as generic "timing component" failure), will they even admit it's an ME thing?


The way this article is written leads me to believe that it is not entirely accurate.


Now this less-mainstream theory about the precarious state of our communication systems has confirmed to a greater degree, would anyone here know of similar risks that few seem to be aware of right now?

I'm not sure if this would be considered OT, but considering the nature and scope of these vulnerabilities I don't consider it reasonable to exclude the possibility of intent and malice.

For this reason I'd like to ask: what do you consider to be "the next, most likely to surface, conspiracy of this flavor"?

The flavor being: "the struggle for control of any and all data and computational resources".


I have a Sun workstation that seems to be no longer supported by Oracle (Sun Ultra 24 with a Q9300). I guess I'll just be vulnerable forever.

I don't really know what AMT does, but this has me thinking, if AMT is provisioned while a machine is used inside a company and then that machine shows up on eBay still provisioned, is it going to be phoning home and still be remotely manageable? How many of these machines have what are essentially persistent rootkits managed by large corporations that have had large fleets of laptops/desktops deployed that are then sold on?


I'm a total n00b to how this stuff works, but I can't seem to find any information for this sort of stuff online. I have an Intel CPU with a Gigabyte Motherboard and BIOS. If I'm running Linux without a GUI (headless) is this something that I have to worry about? If so, how do I turn it off? I don't see any options for the Intel AMT or ME in my BIOS settings.

EDIT: I have a Core i3-4130T. Looks like it doesn't have vPro so I'm hoping I'm safe?


My ignorance is showing, but what product lines are impacted?

Obviously things like Xeons and Core iXs, but what about things like Atom processors in tablets?


The post appears to claim that literally everything is affected, albeit probably only locally exploitable. I think that's what it means at least.


It claims that things with IME are (which I'm not sure if Atom has), and lists a series of architectures of which Atom isn't part. (Its architectures have different names.)

It's ambiguous if the Atom line (and which portions) might be impacted, and I would prefer someone comment directly on if Atom has ME and if so, if it was using the dangerous version (and when).


I think anything with an i5 or i7 in the name has the ME onchip. My spare Thinkpad certainly does & it’s four or five years old at this point in time. I turned the ME off in the bios the moment I acquired it, but I doubt Lenovo will be issuing any bios updates for it.


Intel says Enterprise grade hardware is affected, not consumer grade.


Vulnerable as in how vulnerable? Do you need to be physically connected to local Ethernet for this? WiFi?

If it's WiFi that's damn scary.


Warning: Baseless, Idle Speculation

With the lead time on the silent patch before Shadow Brokers published all the Microsoft exploits, I wonder if Shadow Brokers will be publishing this one soon. No chance of an Intel ME patch going out without being noticed though!

A Shadow Brokers release would be a real mess.


Are remote management functions of portable consumer electronics (i.e.: remotely wiping your iPad) also supported by similar hardware chips from other vendors?


There is a laptop theft recovery/tracking software called LoJack for Laptops (AKA CompuTrace). Some laptop manufactures have added BIOS support for this service (Dell, HP, Lenovo, etc). According to the Wikipedia article [1] this BIOS service copies a downloader into the System32 folder on Windows, which then downloads the full service. It doesn't appear that the BIOS service itself is remotely exploitable, however it can be used for persistent root-kits [2].

[1] https://en.wikipedia.org/wiki/LoJack_for_Laptops [2] https://en.wikipedia.org/wiki/LoJack_for_Laptops#Vulnerabili...


What does this mean?

IPad remote wipe is a function of IOS and the encrypted filesystem it uses on the device, not the CPU.


I've disabled ME on my PC because at some point LMS (Local Management Service) started consuming too much resources for no apparent reason.


How do you disable it? In the BIOS? Is it enabled by default?


I'm running Windows and I just disabled the service. There are a couple of them, one is Local Management Service and the other is User Notification Service.


ME is a separate chip running alongside main CPU. You can't disable it via Windows services :)

I recommend Platform Embedded Security Technology Revealed book [0] from designers and creators of ME for further information.

[0]: https://link.springer.com/book/10.1007/978-1-4302-6572-6


I stand corrected :)


Site is throwing NET::ERR_CERT_AUTHORITY_INVALID on latest Chrome Canary, is anyone else seeing that?


StartCom is untrusted by Chrome these days.


So they (SemiAccurate) knew about this for years, and STILL haven't bothered with disclosure to force Intel's hand earlier?

Thank you, SemiAccurate, for sitting on a vulnerability for years when you could've reported on it long ago and not had us left with this garbage of a security hole to deal with.


A back door is a back door is a back door.

Let's hope Intel and all the other chipmakers will learn this lesson (unless it's done on purpose, in which case they won't care about any lessons learned - they'll do it anyway).


Is there an analog of this issue on AMD chips?


I've always wondering why nobody seems to notice the fact that this site is literally called "Semi Accurate". I mean sure, everyone makes mistake and even the most credible news sources are not entirely accurate all the time. But what am I to think when your organization is literally named after being only half truthful?


It's a semiconductor news site.


Semiconductor Accurate? Doesn't really sound right grammatically, also the arrows missing the target in their logo lead me to believe half accurate was how they intended the name to be interpreted.


The name is a joke. The whole purpose of SemiAccurate is to report leaks and rumors and one can never expect such reporting to be fully accurate.




Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: