KDE KDELibs 4.3.3 Remote Array Overrun (Arbitrary code execution)

2009-11-2000:00:00

vulners.com

JSON

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

[ KDE KDELibs 4.3.3 Remote Array Overrun (Arbitrary code execution) ]

Author: Maksymilian Arciemowicz and sp3x
http://SecurityReason.com
Date:

- Dis.: 07.05.2009
- Pub.: 20.11.2009

CVE: CVE-2009-0689
Risk: High
Remote: Yes

Affected Software:

- KDELibs 4.3.3

NOTE: Prior versions may also be affected.

Original URL:
http://securityreason.com/achievement_securityalert/74

— 0.Description —
KDELibs is a collection of libraries built on top of Qt that provides
frameworks and functionality for developers of KDE-compatible software.
The KDELibs libraries are licensed under LGPL.
— 1. KDE KDELibs 4.3.2 Remote Array Overrun (Arbitrary code execution) —
The main problem exist in dtoa implementation. KDE has a very similar
dtoa algorithm to the BSD, Chrome and Mozilla products. Problem exist
in dtoa.cpp file

http://websvn.kde.org/tags/KDE/4.3.3/kdelibs/kjs/dtoa.cpp?revision=1042584&view=markup

and it is the same like SREASONRES:20090625.

http://securityreason.com/achievement_securityalert/63

but fix for SREASONRES:20090625, used by openbsd was not good.
More information about fix for openbsd and similars SREASONRES:20091030,

http://securityreason.com/achievement_securityalert/69

We can create any number of float, which will overwrite the memory. In
Kmax has defined 15. Functions in dtoa, don't checks Kmax limit, and
it is possible to call 16<= elements of freelist array.

— 2. Proof of Concept (PoC) —

If we use konqueror to see this PoC, konqueror will crash. For example

Program received signal SIGSEGV, Segmentation fault.
[Switching to process 24845, thread 0x7e6e6800]
0x090985c3 in diff () from /usr/local/lib/libkjs.so.5.0

0x06db85c3 <diff+163>: mov %esi,(%ecx)

#0 0x090985c3 in diff () from /usr/local/lib/libkjs.so.5.0
#1 0x0909901b in kjs_strtod () from /usr/local/lib/libkjs.so.5.0
#2 0x090738e5 in KJS::Lexer::lex () from /usr/local/lib/libkjs.so.5.0
#3 0x0907300c in kjsyylex () from /usr/local/lib/libkjs.so.5.0
#4 0x09072f86 in kjsyyparse () from /usr/local/lib/libkjs.so.5.0
#5 0x090805cf in KJS::Parser::parse () from /usr/local/lib/libkjs.so.5.0
#6 0x0908337f in KJS::InterpreterImp::evaluate ()

(gdb) i r
eax 0x0 0
ecx 0x220ff000 571469824
edx 0x0 0
ebx 0x220fbb00 571456256
esp 0xcfbc04e0 0xcfbc04e0
ebp 0xcfbc0518 0xcfbc0518
esi 0xc71c71c7 -954437177
edi 0x0 0
eip 0x21415c3 0x21415c3

esi=0x71c71c7

— 3. SecurityReason Note —

Officialy SREASONRES:20090625 has been detected in:

- OpenBSD
- NetBSD
- FreeBSD
- MacOSX
- Google Chrome
- Mozilla Firefox
- Mozilla Seamonkey
- KDE (example: konqueror)
- Opera
- K-Meleon

This list is not yet closed. US-CERT declared that will inform all vendors about this issue,
however, they did not do it. Even greater confusion caused new CVE number "CVE-2009-1563". Secunia
has informed that this vulnerability was only detected in Mozilla Firefox, but nobody was aware that
the problem affects other products like ( KDE, Chrome ) and it is based on "CVE-2009-0689". After
some time Mozilla Foundation Security Advisory
("http://www.mozilla.org/security/announce/2009/mfsa2009-59.html")
was updated with note :
"The underlying flaw in the dtoa routines used by Mozilla appears to be essentially the same as that
reported against the libc gdtoa routine by Maksymilian Arciemowicz ( CVE-2009-0689)".
This fact ( new CVE number for Firefox Vulnerability )and PoC in javascript (from Secunia), forced
us to official notification all other vendors. We publish all the individual advisories, to formally
show all vulnerable software and to avoid wrong CVE number. We do not see any other way to fix this
issue in all products.

— 4. Fix —
NetBSD fix (optimal):
http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libc/gdtoa/gdtoaimp.h

— 5. Credits —
Discovered by Maksymilian Arciemowicz and sp3x from SecurityReason.com.
— 6. Greets —
Infospec p_e_a pi3
— 7. Contact —
Email:
- cxib {a.t] securityreason [d0t} com
- sp3x {a.t] securityreason [d0t} com