[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <dqiuhc$11a$1@sea.gmane.org>
Date: Tue, 17 Jan 2006 14:21:31 -0000
From: "Dave Korn" <davek_throwaway@...mail.com>
To: bugtraq@...urityfocus.com
Subject: Re: Veritas NetBackup "Volume Manager Daemon" Module Stack Overflow - Exploit
patrickthomassen@...il.com wrote in
news:20060116012440.17540.qmail@...urityfocus.com
> Because the buffer is only very small, I had to write small shellcode.
> The code is less than 100 bytes, and there are 6 bytes left. So there
> is still space to improve it. The stack seems to be static, every run
> at the exact same location.
>
> I used the Import Address Table (that looks like this):
>
> (taken from v5.1)
> Import Address Table
> 00447230 (send)
> 00447234 (recv)
> 00447238 (accept)
> 00447240 (listen)
> 0044724C (connect)
> 00447268 (closesocket)
> 00447284 (bind)
> 00447288 (socket)
>
> Using that shellcode I retrieve the "second" shellcode. This can be ANY
> code, and ANY size. No limitations.
If you want to make your 1st stage _really_ small, you could look further
up the stack in a debugger at the time the overflow hits and see if there's
a local variable there with a handle to the incoming socket on which the
overflow was received. Then you could just recv() more data from it
directly in-line with the overflow packet. Shoudl be able to get it down to
20-30 bytes that way.
> // 'START'
>
> // Move the stackpointer. (0x0012F??? -> 0x0012F000)
> "\xC1\xEC\x0C" // SHR ESP, 0x0C
Right about now would be an "interesting" time to get scheduled /
interrupted / have an APC delivered... !
> "\xC1\xE4\x0C" // SHL ESP, 0x0C
cheers,
DaveK
--
Can't think of a witty .sigline today....
Powered by blists - more mailing lists