full-disclosure - Re: Xfree86 video buffering?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

From: rsw at jfet.org (Riad S. Wahby)
Subject: Re: Xfree86 video buffering?

Stan Bubrouski <stan.bubrouski@...il.com> wrote:
> With this solution someone could intentionally crash your machine to
> avoid those routines from running.  I'm not trying to put you down or
> anything, in fact I probably know less about video related stuff than
> most on the list, this just doesn't seem like the best way to do it.
> I have no better suggestions, I'll leave this one to the experts.

You've found a valid case where clear-on-shutdown breaks, but your
solution doesn't solve the problem, it exacerbates it!

Let's consider the threat here.

- I'm looking at something sensitive on my screen.
- I shut down my computer.
...some time passes...
- Mallory boots my computer and reads out the buffers from my video
  card, thereby obtaining the sensitive information.

By your proposal, we _intentionally_ leave the data on the video card
until Mallory already has access to the machine, trusting that he'll
be stupid enough to load our clear-on-startup video card drivers before
dumping out the framebuffer.

The alternative, to clear-on-shutdown, is obviously better under
normal circumstances, since sensitive data are never left in the
framebuffer. In an exceptional case such as the computer crashing and
failing to clear the framebuffer, it's true that the data remain on the
card. Note, however, that we're in the same situation post-crash that
we'd _always_ be in under your proposed system.

It's obvious that clear-on-startup is strictly worse than clear-on-
shutdown; implementing the latter will prevent a casual observer from
seeing anything sensitive on the screen at startup, but it doesn't
actually grant you any more security, since the data are still
physically on the card.  Implementing both gives you a slight edge
against a passive attacker in a small range of exceptional cases, so it
might be the right way to do things.

Has anyone ever noticed this behavior with Windows, say, after a crash?
I'm not saying I have (haven't actually run a Windows machine regularly
in years, so I probably wouldn't remember anyway), I'm just curious.

-- 
Riad S. Wahby
rsw@...t.org