Re: [Qemu-devel] Minos-type integrity checking in QEMU

[Shashidhar Mysore]

Hi,

Thank you very much for the links. I am looking into your thesis right
now, Congratulations! nice work. I really liked the idea of your
project, especially tracking control dependencies along with data
dependency in QEMU. One question I have is the way in which you obtain
process information from the operating system (The CR3 register,
EPROCESS, etc) - Though you have mainly spoken about Windows, can you
tell me if it is possible to do the same in Linux ? Also, is there a
link from where I can download your tool? or patches to QEMU? 

Thanks again!
-Shashi.

On 4/25/07, maestro <address@hidden> wrote:

Hello Shashi!

you might wanna take a look at the argos [0] project. they do exactly
that kind of stuff.
we implemented something quite similar as the base of my masters thesis
a technical report is available at [1] although what you need is only
covered in a single chapter.

btw. i found it to be easier to make the changes in translate.c instead
of changing every op in op.c

cheers
m.

[0] http://www.few.vu.nl/argos/
[1] http://seclab.tuwien.ac.at/papers/tqana.pdf

Am Mittwoch, den 25.04.2007, 01:16 -0700 schrieb Shashidhar Mysore:
> Hi,
>
> I am trying to incorporate into QEMU (x86) some structures to maintain
> integrity of data that flows in the processor. Specifically, I want to
> maintain a bit for every physical memory address and transfer the
> state information for every address along onto the virtual memory, and
> also down to the instructions which operate on data, and bit arrays
> for registers (so that when an operand is fetched from a particular
> register, we also have the state information for that register). I am
> trying to build a Minos type architecture
> ( http://minos.cs.ucdavis.edu/) but in QEMU instead of Bochs.
>
> For this, as far as I can see, I need to maintain a set of new
> structures to shadow the physical memory and make changes to the micro
> ops defined in target-i386/op.c so that every instruction can track
> the state and store them for every register used as an operand. I'm
> not sure if somebody has already implemented such a system on QEMU? I
> am just beginning to think and work on this, so any
> pointers/suggestions would be greatly appreciated. Please let me know
> your views on this.
>
> Thanks,
> -Shashi.