The concern with the Intel ME is that it has a native network adapter. You can bet efforts are currently underway to discover how to exploit the ME remotely. THAT'S when things get scary.

Your paranoia is not unjustified. Personally, I am nervous that some of my systems have the ME. When attention turned to it about a year ago, i knew it would only be a matter of time before someone broke into it.

Yep, this is the big deal. After I "discovered" the ME, my first stop on my home network was the switch, to block all that crap. (And I found my storage server, equipped with a Supermicro all-in-one motherboard, helpfully grabbed an IP for the ME to listen on with an 'admin/admin' password.)

I just wish the empire builders at the NSA would care about something other than their own little power center. They knew this would happen - it always does. The NSA is probably the biggest security threat to the U.S. people[1] at this point, because they keep building concentrated, high-value targets and then lose control of them.

[1] Not to be confused with 'U.S. government interests'.

This is a useful document for understanding what exactly you're dealing with and what to do about it:

https://www.blackhat.com/docs/us-17/thursday/us-17-Evdokimov...

BMC doesn't always use a dedicated physical port, and it's commonly bridged in sideband to the other NICs on a server.

IOMMU effectively solves the "DMA is completely broken" problem, as far as I'm aware.

Evil Maid attacks are mostly worrisome because even UEFI cannot protect you against some bootloader attacks (what if you disable UEFI or reflash the firmware and then have a bootloader that just looks like a UEFI boot). There are some usages of TPMs that seem quite promising (they revolve around doing a reverse-TOTP-style verification of your laptop to ensure that the TPM has certified the entire boot chain).

It's quite a hard problem, made significantly harder by the fact that every fucking hardware vendor seems to want to make our machines even less secure.

Through this attack, they could compromise the ME longterm, which means the long accepted "nuke it from orbit" solution to security breach (unplug everything, format everything, start from scratch) still wouldn't be enough; that entire chip is done for. And 'using a hack to cleanup the hack' is still in the realm of cleaning up rather than start from scratch, it's not a solution for the same reason than cleaning up your comprised linux box is not one and you need to start from scratch.

https://www.howtogeek.com/56538/how-to-remotely-control-your...

A couple of years back, and being absolutely horrified at the remote management available on my second-hand lenovo t420s - including management over wlan.

Sure the features are gated by price/cpu "brand" - but I think it's safe to assume a) this is complex software and will have bugs with security implications b) once it's well enough understood - it seems likely it can be "upgraded" (similar to how you today can eg: replace the bios with coreboot).

The conclusion is that we need new platforms - perhaps power5 will help.

If the attack can he pulled off with only one time access, it’s worse than an evil maid attack.

I can't help but to think of all those exploits that target anti-virus software.

When you think about it, "it seemed like a good idea at the time" can explain most tragedies in human histories.

So if mass disruption of the very system that support your life can have supporters among a community composed of smart and actively debating people, "it seemed like a good idea at the time" probably happens every week at gov agencies.

What leads to failures, and I suspect happened at Intel, was that they mistook a very localized consensus for a broader one. There's a word for this, it's called "groupthink". A group of people can talk themselves into doing something very stupid (or evil) while still thinking they're doing the right thing, given enough time and motivation.

There was no widespread consensus, outside of Intel, that the IME was a good idea. If they had solicited opinions from outside their organization, they would doubtless have gotten horrified reactions. But they didn't, or if they did they must have dismissed those concerns, because they went through with the bad idea anyway.

The apparent secrecy with which they developed the IME is also a cause for alarm; groups of people who operate in isolation are particularly prone to groupthink, and so even if their motivations are good ones, the fact that they are working without continuous feedback from anyone on the outside raises the chances of a perverse outcome.

At least now that everyone can see the problem people can make informed decisions.

- bribed

- threaten

- felt guilty about it and decided to make amend

My money is that exploits have been on the blacks market for a while now. We just have an official public demo now.

Rule of thumb: when something that catastrophic is made public, the worst already happened and you are late to the party.

In white hats we trust :)

In tin-foil hats, we trust, more like!

In Faraday cages we trust.

And yet...

But yes, I hope that with this it'll be possible to completely remove the remaining few hundred kB of Intel ME code remaining.

[1]: https://github.com/corna/me_cleaner

of course it is, because the USB DCI attack is one level below the Intel ME. Even if it is deactivated via HAP, which basically simply puts the ME code into an infinite loop or a CPU halt state - both can be reversed by JTAG.

If they have full debugger access to what's running in Intel ME then removing the code from the firmware probably doesn't make a difference (assuming they can run un-trusted code in that context). If they cannot write their own code and so an attack requires ROP gadgets then removing the code might make it harder (or impossible) to do, but I doubt it.

[1]: http://www2.lauterbach.com/pdf/directory.pdf#M8.newlink.DIR6...

Huge corporations backed up by gov agencies with a lot of time, money and skilled people VS a few people working for free because they believe they should. Not a fair fight.

I thought this behavior disappeared ~10 years ago.

However there are bugs in USB stacks, especially now you can do so many alternate protocols over USB-C.

https://www.jefftk.com/p/malicious-usb-sticks

There's nothing fundamental in the USB spec that lets devices execute code on the host.

I know some of you might argue that even generic USB sticks can do damage and, whilst I agree, this attack is still a degree worse than most of those.

Thus far the most damage an unknown USB stick could do was type commands as a fake keyboard (visible to user) or exploit a driver vulnerability to silently cause mischief.

Correct me if I'm wrong, but those cases could still be trivially stopped by a security policy set by administrators disallowing unknown USB devices. (Of course, how many places would use this is another matter, but it's still very important for places where this does matter.)

This attack on the other hand would seem to be able to completely bypass your operating systems restrictions on USB ports.

Why? Security knows there are always bugs in software, and assumes they exist. Thanks to @h0t_max, the rest of us know this particular bug exists, but this bug has been around for a while - who's to say evil hax0rs didn't find this bug years ago and have been exploiting it since?

Or put another way - you say an unknown USB stick could exploit a driver vulnerability to silently cause mischief, and then claim that this is easily stopped if an administrator sets a security policy to disallow unknown USB devices. (I assume you mean a Windows GPO enforced policy or similar, and not a written social policy.) Who's to say the code that enforces the policy doesn't have bugs that's exploitable? What if the driver for a known USB stick has exploits?

Physical access is physical access, and while there are mitigations for the evil maid attack (like an encrypted drive and shutting down -not just suspending, when the machine is out of sight), there simply is no way around the fact that physical access is game over.

That mitigation is useless against Evil Maid. There are much more sophisticated mitigations (using a TPM to measure the boot, and then do something akin to TOTP in order to allow the user to actually verify the state of the machine) which actually could protect against Evil Maid almost completely (assuming you don't have something like Intel ME that cannot be verified by the TPM).

"Once you have physical access it's game over" is a very common response to these discussions, and I find it incredibly defeatist. Of course physical access means that the "clock is ticking" until your data is compromised, but sufficient protections can dampen the impact or increase the difficulty.

For example: IOMMU protects against DMA-based attacks, something that was impossible to protect against several years ago. This doesn't mean that someone cannot launch other attacks, but it does mean that the trivial "just plug anything into a USB port and you have DMA" attack is no longer possible.

Encrypted drive + shutdown is a defense against a specific Evil Maid attack, cold boot attacks. It is not a very expensive attack to run; for the cost of a can of compressed air, and a USB drive, anybody sophisticated can run this attack. https://en.wikipedia.org/wiki/Cold_boot_attack

Sorry to sound defeatist, but if you had been relying on IOMMU to save you, the trivial "plug anything into a USB port and you have CPU JTAG access" attack has always been possible. (Never mind that IOMMU implementations aren't guaranteed to be bug free.)

In the face of that, what do you do?

With this knowledge, all I really can do is stay up to date and patch-patch-patch. Have a travel Chromebook for leaving in hotel rooms, but ultimately I just have to know that it's not enough, especially against a CIA-grade Evil Maid, or an Evil Maid that's able to factor 4096-bit prime numbers. (That last one's not theoretical, either. It was revealed a few weeks ago that TPMs in Chromebooks and other hardware was generating weak keys, leading to cloud-factorable 4096-bit RSA keys.)

A less-sophisticated Evil Maid can still physically steal my laptop for pawning, and even if they can't get my data, I've still had my laptop stolen. Not-being-defeatist, I backup my data, although that has a totally different set of security concerns over the Internet.

Access to a USB port is not physical access. USB is a network interface that is commonly used to connect host computers to small portable embedded systems, often tiny NAS units of other people also known as USB flash drives.

This interpretation of "you're being overly paranoid" is new to me ;)

Can you not see the difference between an attacker opening the case and stealing your HD, vs inserting USB key for a few seconds, then going away, and exploiting later at their leisure?

1. https://thehackernews.com/2017/05/intel-amt-vulnerability.ht...

Software do-over is a very well accepted solution (don't bother cleaning the rootkit, just format reinstall), but hardware do-over (change the cpu) is going to be a hard pill to swallow.

https://github.com/openbmc