After the Snowden disclosures -- never mind the whole Room 641A business ( https://en.wikipedia.org/wiki/Room_641A ) -- I think the only safe assumption is bad faith on the part of both government and industry.
In fact, that's the whole problem with extrajudicial domestic spying. In defiance of the principles of both justice and logic, the innocent are presumed guilty, and the burden of proof is shifted to the skeptics.
There isn't anyway to know for sure but its extremely suspicious that they will not release the off switch despite already having developed it for the US government.
Could one write a worm that worked purely on the Intel me, spread machine to machine via me Ethernet monitoring, that could then look at the local hd for crypto keys and report them back to a remote server?
I envision a worm that disables the ME completely once it's found and "infected" another few. Perhaps show a message that says "Your computer is now owned... by you." That would certainly raise some interesting discussion about ethics...
Using which exploit? Quoting the article:
"
The vulnerability identified in CVE-2017-5712 is exploitable remotely over the network in conjunction with a valid administrative Intel® Management Engine credential. The vulnerability is not exploitable if a valid administrative credential is unavailable."
So where is the Ethernet level remote vulnerability? The rest require physical access.
Based on what I've read so far, in my opinion the biggest risk from ME comes from targeted evil maid style scenarios. And I'm not sure if even in those limited conditions ME is reliably exploitable. So I'm more annoyed than panicking right now.
I remove the drivers during deployment. How much does that reduce the vulnerability footprint? Assuming that Lenovo has indeed enabled the firmware write protect feature that Intel describes?
Not installing the drivers prevents your OS from interacting with ME but I don't think it changes the fact that ME has exceptional access to your computer.
As far as I've read, it's more targeted at corporate deployments when the owner/administrator and user aren't the same person or authority. It allows things like remotely reimaging the machine if the OS install gets screwed up, deleting encryption keys if the machine is stolen or otherwise compromised, verifying that the mandatory "endpoint protection suite" is actually running, etc..
You can remotely wake-on-wlan (maybe even 4/5g modem) - remote control, install Os, software - it's a management engine. And it's quite terrifying in its power. But sure, it's potentially useful - it's just a shame it's closed, so it's more of a liability/flaw than a feature.
