PDA

Bekijk Volledige Versie : Re: DCOM RPC exploit (dcom.c)



S G Masood
31/07/03, 14:01
Hello list,


The Dcom.c compiles neatly on Cygwin with GCC 3.2 when
the "#include <error.h>" line is removed.

*Very* accurate. If the machine is vulnerable, the
exploit will almost always succeed on the first
attempt.

I've successfully tested it on about 16 boxes and each
one was rooted on the first try. Among these were
Win2k with SP0, SP1, SP3 while two were WinXP(SP level
not known). Before running the exploit, the machines
were confirmed as vulnerable with the Eeye tool(on a
side note, while the Eeye tool did recognise many
vulnerable boxes, it failed to recognise some of them,
though, they were vulnerable).

One glitch is that the exploitation is not very
stealth. All RPC/COM based functions stop working
completely after exploitation and fail to heal until
the machine is restarted. Many of these functions are
quite visible and easily noticeable(drag&drop,
clipboard, property sheets, etc., for example). This
happens without exception.

The exploit mostly times out when run against remote
hosts.

Hope we are all patched before Tim Mullen's
"Mescaline"(http://securityfocus.com/columnists/174)
becomes a reality.

One last advice - think twice before doing any thing
risky with the exploit. Though highly accurate, it is
very noisy.


Regards,

S.G.Masood

Hyderabad,
India.

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com

S G Masood
31/07/03, 14:01
Hello list,


The Dcom.c compiles neatly on Cygwin with GCC 3.2 when
the "#include <error.h>" line is removed.

*Very* accurate. If the machine is vulnerable, the
exploit will almost always succeed on the first
attempt.

I've successfully tested it on about 16 boxes and each
one was rooted on the first try. Among these were
Win2k with SP0, SP1, SP3 while two were WinXP(SP level
not known). Before running the exploit, the machines
were confirmed as vulnerable with the Eeye tool(on a
side note, while the Eeye tool did recognise many
vulnerable boxes, it failed to recognise some of them,
though, they were vulnerable).

One glitch is that the exploitation is not very
stealth. All RPC/COM based functions stop working
completely after exploitation and fail to heal until
the machine is restarted. Many of these functions are
quite visible and easily noticeable(drag&drop,
clipboard, property sheets, etc., for example). This
happens without exception.

The exploit mostly times out when run against remote
hosts.

Hope we are all patched before Tim Mullen's
"Mescaline"(http://securityfocus.com/columnists/174)
becomes a reality.

One last advice - think twice before doing any thing
risky with the exploit. Though highly accurate, it is
very noisy.


Regards,

S.G.Masood

Hyderabad,
India.

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com

31/07/03, 14:01
In-Reply-To: <20030727025321.64988.qmail@web11001.mail.yahoo.com>

>One glitch is that the exploitation is not very
>stealth. All RPC/COM based functions stop working
>completely after exploitation and fail to heal until
>the machine is restarted. Many of these functions are
>quite visible and easily noticeable(drag&drop,
>clipboard, property sheets, etc., for example). This
>happens without exception.

If the shellcode exit via ExitThread(), RPCSS will not die, everything
rock as usual, and you can run the exploit over and over again.

sk

31/07/03, 14:01
In-Reply-To: <20030727025321.64988.qmail@web11001.mail.yahoo.com>

>One glitch is that the exploitation is not very
>stealth. All RPC/COM based functions stop working
>completely after exploitation and fail to heal until
>the machine is restarted. Many of these functions are
>quite visible and easily noticeable(drag&drop,
>clipboard, property sheets, etc., for example). This
>happens without exception.

If the shellcode exit via ExitThread(), RPCSS will not die, everything
rock as usual, and you can run the exploit over and over again.

sk

Martin Peikert
31/07/03, 14:01
Hello,

S G Masood wrote:
> The Dcom.c compiles neatly on Cygwin with GCC 3.2 when
> the "#include <error.h>" line is removed.

s/error.h/errno.h/ and dcom.c will compile on cygwin/gcc3.2.

GTi

Martin Peikert
31/07/03, 14:01
Hello,

S G Masood wrote:
> The Dcom.c compiles neatly on Cygwin with GCC 3.2 when
> the "#include <error.h>" line is removed.

s/error.h/errno.h/ and dcom.c will compile on cygwin/gcc3.2.

GTi