Re: [gentoo-user] New Intel vulnerability?

[Michael]

On Friday, 6 March 2020 13:48:00 GMT Rich Freeman wrote:
> On Fri, Mar 6, 2020 at 3:50 AM Michael <confabulate@××××××××.com> wrote:
> > I have lost count with the naming scheme of Intel's embedded spyware to
> > know if this is yet another vulnerability, or something to convince me to
> > throw away the last Intel powered box still in my possession (mind you
> > its >10yr old):
> >
> > https://arstechnica.com/information-technology/2020/03/5-years-of-intel-cp
> > us-and-chipsets-have-a-concerning-flaw-thats-unfixable/
> The article is actually pretty well-written.  I haven't studied the
> issue in any depth but here are my impressions:
>
> 1.  You need a firmware update to prevent software vulnerabilities.
> 2.  Even with a firmware update you are vulnerable to somebody with
> physical access to your device.
>
> The whole issue centers around TPM essentially.  This potentially
> impacts you if you don't care about TPM, but it impacts you more if
> you do care about TPM.
>
> If you don't use TPM (probably many on this list), then your main
> concern should just be with getting your firmware patched (#1 above).
> Otherwise you could be vulnerable to rootkits that hijack the TPM on
> your machine and use it to spy on you in hard-to-detect ways.  Based
> on the article a firmware patch should block the ability for software
> to get into your TPM and mess with it.  Then you're basically safe.
> If you aren't using TPM you're already vulnerable to somebody with
> physical access to your device, so there is no real change in the
> threat model for you.
>
> Now let's get to those who use TPM or the other impacted trusted
> services.  You use these if:
> 1.  You rely on secure boot (with any OS - Linux does support this
> though I imagine it is rare for Gentoo users to use it).
> 2.  You rely on TPM-backed full disk encryption.  This includes
> Bitlocker and most commercial solutions.  This doesn't include LUKS.
> If your disk is unreadable if you remove it from the computer, but you
> don't need any password to boot it, then you're probably using
> TPM-backed encryption.
> 3.  You are Netflix/etc and are relying on remote attestation or any
> of the technologies RMS would term "treacherous computing."
> 4.  You are a corporate owner of computers and are relying on the same
> technologies in #3 but to actually protect your own hardware.  Or
> maybe if you're the only person in the world using Trusted GRUB.
>
> If you fall into this camp you need to still update your firmware to
> address the non-TPM-user and to avoid making it trivial for software
> to steal your keys/etc.  However, you need to be aware that you are no
> longer secure against physical theft of your device.  Somebody who
> steals your laptop with passwordless encryption might be able to break
> the encryption on your device.  They would need to steal the entire
> laptop though - if you throw out a hard drive nobody will be able to
> recover it from the trash.  If you're Netflix I'm not sure why you're
> even bothering with this stuff because all your content is already
> available in full quality on torrent sites, but I guess you can lose
> even more sleep over it if you want to.  If you're using secure boot
> then somebody with physical access might be able to change the
> authorization settings and let another OS boot.  If you're a
> corporation with sensitive data you probably have the biggest impact,
> because you're distributing laptops to people who lose them and who
> don't have a ton of security hygiene to begin with.
>
> The only people who probably will consider replacing hardware are
> corporate users.  Most on this list are going to be fine with a
> firmware update as you're probably not using the TPM features.
> Indeed, even getting those working on Linux is a PITA - I'm not aware
> of any distro that has TPM-backed encryption out of the box.  Windows
> has this in the pro edition (Bitlocker) and it is probably fairly
> popular.
>
> If you use LUKS-based encryption you are going to be secure with
> patched firmware as long as nobody installs a keylogger on your
> device.  That will be easier with the vulnerability, though somebody
> could just hack the keyboard hardware anyway and LUKS wouldn't protect
> you against that.  TPM has pros and cons compared to LUKS in general.
> If you don't patch your firmware then it is possible a rootkit might
> get in there and steal your keys at boot time.
>
> If somebody has more to add from researching this more I'm all ears.
> Now I need to check if my windows tablet with Bitlocker is vulnerable.
> This also shows the downside to TPM encryption - it is convenient but
> if somebody steals a laptop and just keeps it stored away they could
> always use a vulnerability like this to break in sometime in the
> future.  It is probably still worth using as a minimum because it does
> protect against hard drive loss, and it works if your TPM isn't
> vulnerable.

Thanks for this analysis Rich, quite thorough as usual.

TBH I have avoided using TPM so far because it requires an implicit trust on 
the OEM and most observers and reports evidence this is invariably misplaced.

I seem to recall a TPM vulnerability (not sure which version, I think TPM2), 
which cause TPM to always spew out the same limited number of ephemeral keys, 
making an unwarranted entry by a determined attacker possible.

This is another reason I don't trust obscure closed code solutions like 
Ubikey.