Researchers Hack Intel's VPro

snydeq writes "Security researchers from Invisible Things Lab have created software that can 'compromise the integrity' of software loaded using Intel's vPro Trusted Execution Technology, which is supposed to help protect software from being seen or tampered with by other programs on the machine. The researchers say they have created a two-stage attack, with the first stage exploiting a bug in Intel's system software. The second stage relies on a design flaw in the TXT technology itself (PDF). The researchers plan to give more details on their work at the Black Hat DC security conference next month."

Comments

Re:TXT? PDF? Wha?

[CannonballHead]

My mistake.

1. can't tell if you're joking.

2. Execution, not Execution.

Thank you!

[Just Some Guy]

RMS calls this "treacherous computing", and I have to agree with him. This is a good development as it demonstrates quite nicely that DRM (which is probably the #1 use of VPro et al) in simply not possible. Thanks, ITL, for showing this as folly!

Re:Thank you!

[IamTheRealMike]

Keyword, at Intel. TC is the work of a large committee, with many companies. If you read the specs the conflicting goals are obvious. Simple question - is the TPM meant to resist hardware attacks or not? Sometimes it is, sometimes it isn't. It's not very good at this currently, you could beat 1.1 TPMs with a piece of wire (literally), but Intel are moving them inside the south bridge, where hardware attacks will be much harder.

In theory at least TC can be used to implement better DRM, because it makes it harder for people to debug the implementation. But there are still many unimplemented features needed to make this work, eg, trusted I/O, and no real roadmap to implement them. And even when done, it'll be years before the technology is widespread, and it's so complicated I'm sure Joanna and friends will be able to find many more problems with it.

The real promise of TC is a way out of the malware quagmire. Being able to run a web browser and know - for sure - that it's not been compromised by a password sniffer or the like, well, that's a useful thing and that's what TXT lets you do (when complete). A remote voting app that can prove to the server that it's a real human casting the vote and not a bot? A very useful thing, perhaps even a necessary precondition for digital democracy. TC can make this happen. DRM? Well if you want a crappy inferior very complex form of DRM then sure, go ahead, but it'll be less secure and more expensive than the equivalent implemented in controlled hardware like the PS3, Xbox360, mobile phones etc ...

Re:Thank you!

[Just Some Guy]

Bullshit, not a single person working on TPM at Intel thinks it will ever work for DRM.

Funny, as it's the first listed possible application on Wikipedia. How could TPM possibly not be used for DRM? All the ingredients are there. From the same article:

Sealed storage could prevent users from moving sealed files to the new computer. This limitation might exist either through poor software design or deliberate limitations placed by publishers of works. The migration section of the TPM specification requires that it be impossible to move certain kinds of files except to a computer with the identical make and model of security chip.

Isn't that almost the very definition of DRM?

Re:Thank you!

[Anonymous Coward]

Excuse me... let me phrase that correctly: "Bullshit, not a single person working on TPM at Intel will admit that was designed for DRM."

The entire reason for the project (started back in the late 90s) was DRM - or, as one Intel engineer at a talk I attended put it - "making a system secure against its owner". Only later they decided, after users started to realise just what TXT really means for them (total control by the likes of Microsoft), that they would smother the whole "for DRM" thing and flatly refuse to ever discuss it. Instead they always emphasise the "security" aspects instead. Only morons are fooled - hello there.

Anyone who thinks that Intel is not about DRM is an idiot. Intel is *THE* DRM kingpin (HDCP etc etc).

Re:Thank you!

[Alsee]

Orly?

What a load of crap. At best you are merely naive.

I am a programmer, and in particular I have studied the Trusted Platform Technical Specification documentation. All 332 pages of dense technicaleese. There is one particular page I would like to cite. In the TCPA Main TCG Architecture v1_1b.pdf on page 277 the documentation comes right out and announces the fact it is designed to be secure against "rogue Owners".

You are either mistaken, or you're full of crap. The chip is in fact designed to lock the computer against the owner. Yes, locks that are designed to protect the computer against it's owner will also prevent outside attackers from doing things that the owner himself is forbidden to do. However that is incidental. A hostile Trusted Computing system trying to lock computers against their owners is fundamentally different than a system designed to secure computers for the owner.

If you really do believe that this is solely intended for the benefit of the owner, perhaps you could answer some questions for me.

Why the absolute refusal to implement the EFF's Owner Override proposal? It would give the owner full control of his own computer while still securing against remote attacks. You could even secure against local attackers (other than the owner) by placing adding some sort of Owner Authentication element to the Override system.

Or how about my proposal? I merely want a printed copy of the master key to my own computer. I merely want the option to buy a computer that comes with a printed copy of my master key. (Technical note: I am referring to the PrivEK key, and having the option to export the RSK key encrypted to the PrivEK would be beneficial for ease and security reasons.) Go ahead, explain to why I am absolutely forbidden to know the master key to my own computer. Go ahead and explain why they absolutely refuse to PERMIT anyone to manufacture any compatible Trust Chip that permits the owner to know their own master key.

And best of all, explain to me all of the documented systems and plans to REVOKE and (for all practical purposes) brick any chip if they ever detect that you have learned the master key locked inside you computer, if you ever learn the master key to control your own computer, if they ever detect that you have the power and control to override any DRM system based on the chip.

And don't even try the line about how this revocation system is "not part of the chip itself". The chip was explicitly designed to secure the computer against the owner, the chip was explicitly designed to to support that revocation system, and the chip's technical documentation and design specification explicitly mention this revocation system.

The design specs endlessly list all of the things that the owner MUST be forbidden to be able to do, all of the things the owner MUST be forbidden to know, the specification even has a section that mandates that any owner's data under "non-migable keys" MUST be effectively impossible to back up and MUST be irretrievably lost if the chip ever dies.

And on and on and on. Yes, the chip was explicitly designed to consider the owner to be the enemy. The chip is explicitly designed to be secure against "attacks" by the owner. Yes, the current generation of chips are relatively vulnerable to physical attack - by the owner or by a hostile attacker. However it is fundamentally designed to lock against the owner, there is a supplemental specification on how to increase the physical security against the owner and how to certify hardware as possessing stronger anti-owner physical security, and there is mention in the CHIP speck itself and in supplemental specifications on how to revoke and lock-out any chip where an owner does manage to gain local override control over his own computer.

Yes, there are some people working on Trusted Computing with the intent of securing your computer for you, of protecting you against remote attackers. However that does not change the fact tha

Re:Thank you!

[jhol13]

Being able to run a web browser and know - for sure - that it's not been compromised by a password sniffer or the like, well, that's a useful thing and that's what TXT lets you do (when complete).

No it won't. If the said browser behaves erroneously on a particularly crafted web page the web page creator might be able (depending on the error) to take full control of the machine, e.g. by injecting remotely controllable ("telnet") Javascript applet.

For voting the TC cannot *prove* anything - again a simple overflow (either buffer or integer or ...) bug can make the bot look exactly like human to the TC. TC can "prove" provided there are no bugs. Which is lame.

Another repeat: the unlockable lock

[Anonymous Coward]

Never a lock has been created that can't be broken.

Any time you see "unbreakable", "unsinkable" or similar claims, call your bookie: they will. The question is when, not if.

Re:Another repeat: the unlockable lock

[nobodylocalhost]

risking to be modded troll, i would like to say sure there is an unbreakable lock. An unbreakable lock is a lock that noone cares enough to break.

Re:Wii Homebrew Channel

[whoever57]

The Wii has perfect encryption and signing on hardware-assisting firmware and system software that can't be compromised.

Let me correct that for you:

The Wii has perfect ^H^H^H^H^H^H an encryption and signing on hardware-assisting firmware and system software that can't be ^H^H^H^H^H^H hasn't been compromised.

Wrong Wrong Wrong

[Glasswire]

vPro is mostly about AMT OOB management which is secure and is in it's 5th generation. TXT is relatively new component which is implemented virtually nowhere yet and has virtually nothing to do with the AMT functionality that has been and is being implemented hundreds of sites. AMT management is 97% of what vPro really is and is what the industry system OEMs generally mean when they say vPro. TXT is a future technology waiting for ISV enablement whereas core AMT/vPro is real and here now. Saying that because TXT may be compromised AND suggesting that the primary, working part of vPro is insecure is outrageously misleading.