OPENSSH REMOTE ROOT COMPROMISE ALL VERSIONS (fwd)
Pekka Savola
pekkas at netcore.fi
Thu Jan 9 00:54:55 EST 2003
There has not been any mention of this, so..
I'm having hard time trying to figure out whether this is for real or just
a joke. (Anyway, I don't use those possibly vulnerable features myself.)
Any analysis yet?
--
Pekka Savola "You each name yourselves king, yet the
Netcore Oy kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings
---------- Forwarded message ----------
Date: Sat, 4 Jan 2003 19:37:03 -0800
From: mmhs at hushmail.com
To: bugtraq at securityfocus.com
Subject: OPENSSH REMOTE ROOT COMPROMISE ALL VERSIONS
-----BEGIN PGP SIGNED MESSAGE-----
*********** OPENSSH REMOTE ROOT COMPROMISE ALL VERSIONS ***********
MICKEY MOUSE HACKING SQUADRON ADVISORY #2
DISCLAIMER
- ----------
The nation's zeroth private security intelligence firm, Mickey Mouse
Hacking Squadron uniquely addresses the challenges faced by both public-
and private-sector organizations in protecting critical information
assets.
Our intelligence is timely, delivered 24 x 7, 365 (*) days per year;
relevant, fully customizable, and actionable intelligence is only
valuable if it makes a difference.
(*) in the case of a leap year, we of course provide a 24 x 7, 366 days
premier service.
TECHNICAL BACKGROUND
- --------------------
The following advisory is based on the excellent advisory published by
Global InterSec LLC *six months ago*:
http://www.globalintersec.com/adv/openssh-2002062801.txt
After more than six months of intensive underground research, our ISO
31337 certified security department evidenced that the bug (an integer
overflow, resulting in a heap overflow) described in the aforementioned
advisory still exists in OpenSSH 3.5p1 and 3.4p1, and remains trivially
exploitable. All existing PAM enabled versions of OpenSSH (3.5p1, 3.4p1
and below) are therefore affected.
Due to various advisories posted to various fora by unnamed security
companies, this bug was supposed to be nonexistent or nonexploitable.
Fortunately, Global InterSec LLC shed some light on the whole affair and
revealed the malignant nature of the oversight to the world.
Their results were applied to the latest OpenSSH versions by privately
trained Mickey Mouse Hacking Squadron security specialists and revealed
that the exploitation techniques developed by Global InterSec LLC are
still applicable to the newest OpenSSH.
PROOF OF CONCEPT
- ----------------
The following proof of concept is reproducing Global InterSec LLC
findings, enhanced with the patented research performed by Mickey Mouse
Hacking Squadron against OpenSSH 3.5p1.
First of all, the OpenSSH 3.5p1 server has to be built (with PAM support
enabled):
$ tar xzf openssh-3.5p1.tar.gz
$ cd openssh-3.5p1
$ configure --with-pam
[...]
$ make sshd
[...]
Before the SSH server is actually executed, the sshd_config file should
be modified in order to enable PAM ("PAMAuthenticationViaKbdInt yes").
# sshd
In order to reveal the nature of the OpenSSH vulnerability, the next
step is to connect to the SSH server:
$ ssh werewolf.research.mmhs.com
Password:
Thanks to the "Password:" prompt, it is clear that PAM is actually
enabled (otherwise, the prompt would have been "user at host's password:").
This unique fingerprinting technique was investigated by Mickey Mouse
Hacking Squadron, and is already present in the latest version of the
Mickey Mouse Hacking Squadron award winning network vulnerability
assessment tool.
After the previous command was executed, the freshly spawned sshd
process has to be examined with a debugger, in order to set the correct
breakpoints within the input_userauth_info_response_pam() function of
OpenSSH, as demonstrated in the Global InterSec LLC advisory:
# gdb sshd 6552
(gdb) disassemble input_userauth_info_response_pam
[...]
0x80531bc <input_userauth_info_response_pam+192>: push %esi
0x80531bd <input_userauth_info_response_pam+193>:
call 0x807306c <xfree>
[...]
(gdb) break *0x80531bd
Breakpoint 1 at 0x80531bd: file auth2-pam.c, line 158.
(gdb) continue
Continuing.
Now that the buggy call to xfree() can be intercepted, the SSH client
should trigger the integer overlow and the resulting heap overflow:
$ ssh werewolf.research.mmhs.com
Password: <type a thousand 'A' characters here and hit enter>
After that, the xfree() breakpoint is reached, and the next call to
free() should therefore be intercepted in order to comply with the
technique developed by Global InterSec LLC:
Breakpoint 1, 0x080531bd in input_userauth_info_response_pam (type=61,
seqnr=7, ctxt=0x809c050) at auth2-pam.c:158
158 xfree(resp);
(gdb) disassemble xfree
[...]
0x807308e <xfree+34>: call 0x804ba14 <free>
[...]
(gdb) break *0x807308e
Breakpoint 2 at 0x807308e: file xmalloc.c, line 55.
(gdb) continue
Continuing.
Breakpoint 2, 0x0807308e in xfree (ptr=0x809dfb8) at xmalloc.c:55
55 free(ptr);
(gdb) x /10x 0x809dfb8
0x809dfb8: 0x41414141 0x41414141 0x41414141 0x41414141
0x809dfc8: 0x41414141 0x41414141 0x41414141 0x41414141
0x809dfd8: 0x41414141 0x41414141
>From here on, as demonstrated by Global InterSec LLC, exploitation
becomes trivial. For more information on exploiting calls to free() see
the excellent Phrack article "Once upon a free()" [2].
WORK AROUND
- -----------
As mentioned in http://www.openssh.com/txt/preauth.adv, and as
demonstrated by noir in http://www.phrack.org/phrack/60/p60-0x06.txt,
"you can prevent privilege escalation if you enable
UsePrivilegeSeparation in sshd_config."
Love,
- --
Mickey Mouse Hacking Squadron
-----BEGIN PGP SIGNATURE-----
Version: Hush 2.2 (Java)
Note: This signature can be verified at https://www.hushtools.com/verify
wlkEARECABkFAj4XqFwSHG1taHNAaHVzaG1haWwuY29tAAoJEMZ9fu0iAPxbgYEAoL0W
0oGQQvqwwZAGADonQ2TOUjNmAJ4zuUfANSpju97UjXdD65bkCy6M1A==
=YvOU
-----END PGP SIGNATURE-----
Concerned about your privacy? Follow this link to get
FREE encrypted email: https://www.hushmail.com/?l=2
Big $$$ to be made with the HushMail Affiliate Program:
https://www.hushmail.com/about.php?subloc=affiliate&l=427
More information about the openssh-unix-dev
mailing list