On Thursday, September 7, 2017 at 12:49:05 PM UTC-4, cooloutac wrote:
> On Thursday, September 7, 2017 at 9:40:47 AM UTC-4, Patrik Hagara wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA256
> > 
> > On 09/07/2017 03:21 PM, cooloutac wrote:
> > > On Tuesday, August 29, 2017 at 12:46:04 PM UTC-4, Patrik Hagara
> > > wrote: On 08/29/2017 06:25 PM, cooloutac wrote:
> > >>>> On Tuesday, August 29, 2017 at 11:38:59 AM UTC-4, Patrik
> > >>>> Hagara wrote: On 08/29/2017 04:50 PM,
> > >>>> [email protected] wrote:
> > >>>>>>> Leo Gaspard,
> > >>>>>>> 
> > >>>>>>> I have read about AEM but have never used it, it seems
> > >>>>>>> like it is geared towards protecting from USB's with
> > >>>>>>> malicious firmware on them.
> > >>>>>>> 
> > >>>>>>> Does AEM actually verify the integrity of /boot before
> > >>>>>>> using? This is what I am looking for, either a method
> > >>>>>>> of encrypting /boot or even better, a secure method to
> > >>>>>>> verify its integrity during boot
> > >>>>>>> 
> > >>>> 
> > >>>> AEM does verify the integrity of /boot using the TPM
> > >>>> seal/unseal operation. If any of the boot components change,
> > >>>> AEM falls back to regular, unmeasured boot -- the user is
> > >>>> expected to notice this and cease using the
> > >>>> potentially-compromised system (the lack of explicit
> > >>>> indication of failed AEM boot is deliberate, see the last FAQ
> > >>>> item at [1]).
> > >>>> 
> > >>>> The security provided by AEM is much more stronger than
> > >>>> encrypted /boot could ever offer, because as already pointed
> > >>>> out, there is always a little first-stage bootloader stub on
> > >>>> the disk unencrypted and it would be easy to overwrite it
> > >>>> with a malicious version that would intercept the encryption
> > >>>> passphrase and exfiltrate it using eg. unused disk sectors.
> > >>>> 
> > >>>> If someone did the above with AEM, the TPM would refuse to
> > >>>> useal the AEM secret as the platform state would be
> > >>>> different.
> > >>>> 
> > >>>> The feature protecting dom0 from malicious USB devices [2]
> > >>>> serves a different purpose and is not related to AEM. Plus,
> > >>>> you always need to disconnect all untrusted USB devices while
> > >>>> rebooting Qubes, regardless of whether you have USB qube set
> > >>>> up or not.
> > >>>> 
> > >>>> 
> > >>>> Cheers, Patrik
> > >>>> 
> > >>>> 
> > >>>> [1] 
> > >>>> https://blog.invisiblethings.org/2011/09/07/anti-evil-maid.html
> > >>>> [2] https://www.qubes-os.org/doc/usb/
> > >>>> 
> > >>>> my problem is unfortunately i can't keep buying new pc's. SO
> > >>>> maybe its all for naught for me.. Also AEM does not seem as
> > >>>> reliable as secureboot. Alot of issues on threads with some
> > >>>> people. It seems complicated. even false alarms. But I really
> > >>>> do think they are supposed to compliment each other.  trusted
> > >>>> boot and measured boot yes? AEM definitely comes in handy for
> > >>>> people who would find it nescessary to buy new equipment.
> > > 
> > > Well, it depends... If you can be reasonably sure that the
> > > attacker did not modify any firmware (eg. the network card), you
> > > could simply reinstall Qubes from a trusted install media and
> > > restore VMs from a backup. This becomes trivial once stateless
> > > computers [1] are a thing.
> > > 
> > >>>> But I would still want an encrypted boot, if I was going to
> > >>>> use aem,  and key on a external usb disk,  which
> > >>>> unfortunately means I could not use a sys-usb.  Am I wrong
> > >>>> about this?  Does this change in 4.0?
> > > 
> > > It is possible to use AEM along with USB qube, you just have to 
> > > disable the early hiding of USB devices from dom0. But once Qubes
> > > is fully booted and sys-usb started, you get the full USB qubes
> > > protecion.
> > > 
> > >>>> So this is where the what fits into you "security model"
> > >>>> What are you more concered about. I assumed we had to choose
> > >>>> between aem vs sys-usb?  For people travelling with laptops
> > >>>> in strange places where physical compromise is more likely
> > >>>> aem is a good option. Some people would prolly not just buy a
> > >>>> new laptop but know when to destroy theirs too lol.
> > > 
> > > The only trade-off for AEM regarding USB devices is that the USB
> > > stick you buy to install AEM on could be compromised already
> > > (straight from the factory, or intercepted & infected during
> > > shipping). And since you need to plug it directly into dom0 in
> > > order to perform the install, it could exploit USB or filesystem
> > > drivers and gain control of dom0.
> > > 
> > > So it is kind of a trust-on-first-use scenario for the
> > > installation step only.
> > > 
> > > 
> > > Cheers, Patrik
> > > 
> > > 
> > > [1] https://blog.invisiblethings.org/papers/2015/state_harmful.pdf
> > > 
> > > Doesn't Qubes assume the netcard will be compromised, hence
> > > untrusted.
> > 
> > It does, but with compromised firmware the NIC can inject code into
> > the OS during early boot (before the NIC is isolated from DMA using
> > the IOMMU). AEM will catch/prevent this.
> > 
> > > So good to know we can use a usb key with a sys-usb.   IS there
> > > some sort of risk to doing this?  Do we have to pull out the usb
> > > stick quickly after booting before system loads or something?
> > 
> > Not sure about the currently available AEM version, but if/when my
> > changes to AEM get merged, you get prompted to remove the AEM USB
> > stick before Qubes will finish booting (so that the untrusted sys-usb
> > cannot read the contents of the stick).
> > 
> > In case the available version does not do this, one possible
> > work-around is disabling sys-usb VM auto-start and start it manually
> > after unplugging the AEM stick. This should IMO be the recommended
> > thing to do anyway, as otherwise it is difficult and potentially
> > insecure to update AEM/Xen/kernel stored on the stick, as you need to
> > shut down sys-usb and manually re-assign the USB controller to dom0,
> > which, if sys-usb was compromised and somehow altered the USB
> > controller incapable of doing a proper PCI Function Level Reset (AFAIK
> > this HW feature is commonly broken, especially for GPUs), it could
> > gain control of dom0.
> > 
> > 
> > Cheers,
> > Patrik
> > -----BEGIN PGP SIGNATURE-----
> > Version: GnuPG v2
> > 
> > iQIcBAEBCAAGBQJZsUxSAAoJEFwecd8DH5rlpPQP/iV6UzUWh2W7z9R/7vQd9qh+
> > TMJTRJkczaBC38i/yNvN2vTDst7EArQYyViyKL1gSdTlbkkaDy1gvg3loav0Nc3I
> > a6Bqnp046uY04152QFCaONFw/TyfN5eLLUYrLfKwtB5TIEVY7bBRkuMkbTnJO+DB
> > VzTdtKY6vCAaFm30e8GDhj/9bfwTz0Y8hPTpJTCZ37BcYl52OjPhQ3VsrRjVorf8
> > 3JtfxlJXMuU3Ap6vaM+HdhoJKKjraAasNYlwuJXjPWzBpuWYWEmMFohfkCdFr1N/
> > Gli4q4zS/QUHChxCSTHY5q4SOfAGWcBugGlWpW0GEgbi+bmXB4SAsFTmMDhg1nXo
> > A85n0/idbJU5d7xOzTrfaA+uDAXueAktnACjTZ9CTplMGv0hBGOY9dp8N3a+UTHr
> > dipgjZDpc+FByZLR1wZ89899T+lDDnzCU38/knhCsmlnxdhCd1mzspjWIir/Ngfs
> > C/WKc5P4gHFEJNQmsU3lEAFX80dvG7NcYo8zEwppGYFQY8afpepvsifLHl8YnEdX
> > xuR9BRKd626KiHEAPxXAHjCfJOkM+l3Oyq+hg8GDvJft6YjTarfPA2c/52y7hZej
> > UW/pYuKJbK+Nkwsqxu8p6I1qf/MtWf6HsI+m5+e33FKnu/qY+oA+j1ETM1aOnN1I
> > BYXLyrZLwh+4dhWknEU8
> > =RDNz
> > -----END PGP SIGNATURE-----
> 
> only problem with remembering to start the sys-usb is we all forget sometimes.

though maybe for the next final release I will try aem too tks for the info.   
Although probably should only be doing it once I buy new hardwar. Its probably 
already too late for me lol,   but i'm sure other changes could potentially 
happen as well. 

-- 
You received this message because you are subscribed to the Google Groups 
"qubes-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/qubes-users/1fc5b701-46f3-4822-845d-f8705e4e32a6%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to