Sunday, December 10, 2017

BlackHat Europe 2017 : THE GREAT ESCAPES OF VMWARE

This December 2017 Me & my colleague Yakun Zhang delivered a talk at Blackhat Europe 2017 Briefings on VMWare escapes. Blackhat Europe is an annual information security conference, scheduled on December 4 2017 to December 7, 2017, in the ExCeL London, located at 1 Western Gateway, London E16 1XL.We have talked about reverse engineering of vmware, attacking hypervisor isolation and some virtual machine escape attacks against vmware.


Talk Abstract

Virtual machine escape is the process of breaking out of the virtual machine and interacting with the host operating system. VMWare recently fixed several bugs in their products that were allowing malicious code to escape sandbox. Some of these issues were exploited and reported during exploitation contest and while others reported individually by researchers. For very obvious reason details of this bugs are undisclosed. This paper presents a case study of VMWare VM escape vulnerabilities based on the analysis of different patches released by VMWare in recent past. 

Looking at the advisories published by VMWare in the last few months, reveals that there are many surfaces, that are being targeted by security researchers. To summarize, the attack surfaces would be as follows: 

A) RPC Request handler.
B) Virtual Printer.
C) VMWare Graphics Implementation.

Talking about vulnerabilities fixed in VMWare RPC layer, we see several CVEs (CVE-2017-4901, CVE-2016-7461 etc.) fixing security issues in RPC layers. This talk will cover end to end RPC implementation in VMWare workstation. It will cover everything from VMWare Backdoor in guest OS to different RPC command handler in host OS. We will uncover some of these fixed bugs in VMWare RPC layer by performing binary diffing on VMWare Workstation binaries. This talk will also showcase some of the PoCs developed from different VMware workstation patches.


VMWare's EMF file handler is one of most popular attack surfaces, when it comes to guest to host escape. VMSA-2016-0014 fixed several security issues in EMF file handling mechanism. EMF format is composed of many EMR data structures. TPView.dll parses every EMR structure in EMF file. In VMware, COM1 port is used by Guest to interact with Host printing proxy. EMF files are spool file format used in printing by windows. When a printing EMF file request comes from Guest, in host TPView.dll render the printing page. The TPView.dll holds the actual code which parses the EMF file structures. In our talk, we will be diving deep into this attack surface & uncover some of the vulnerabilities fixed in this area recently by performing binary diffing on VMWare work station binaries.


VMSA-2017-0006 resolved several security vulnerabilities in Workstation, Fusion graphics implementation which allows Guest to Host Escape. These vulnerabilities were mostly present in VMWare SVGA implementation. In this section of our talk we will cover implementation of VMWare virtual GPU through reverse engineering different guest components (vmx_fb.dll - VMware SVGA II Display Driver, vmx_svga.sys - VMware SVGA II Miniport) as well as host component (vmware-vmx.exe) where virtualize GPU code exist. The VMware virtual GPU provides several memory ranges which is used by Guest OS to communicate with the emulated device. These memory ranges are 2D frame buffer and FIFO Memory Queue. In FIFO memory queue, we write command that we want our GPU to process. The way VMWare handles and process these commands is error prone. This talk will uncover some of these bugs in SVGA command processing code and try to understand anatomy of issues by bin-diffing through VMWare binaries.

Slides:


References:


  • https://www.blackhat.com/eu-17/briefings.html#the-great-escapes-of-vmware-a-retrospective-case-study-of-vmware-g2h-escape-vulnerabilities
  • https://www.blackhat.com/docs/eu-17/materials/eu-17-Mandal-The-Great-Escapes-Of-Vmware-A-Retrospective-Case-Study-Of-Vmware-G2H-Escape-Vulnerabilities.pdf



Sunday, October 8, 2017

BruCON 2017: Browser Exploits? Grab ’em by the Collar!

This year in October I have delivered a talk about reliable detection of Browser Exploit at Brucon 2017. BruCON is an annual security and hacker(*) conference providing two days of an interesting atmosphere for open discussions of critical infosec issues, privacy, information technology and its cultural/technical implications on society. Organized in Belgium, BruCON offers a high quality line up of speakers, security challenges and interesting workshops.


Talk Abstract:


APT has become a hot topic in enterprise IT today. One of the softwares that we see becomes victim of APT attack more often is web browsers and the attack surface is becoming bigger and bigger every day.

TCP Live Stream Injection (https://en.wikipedia.org/wiki/Packet_injection) is a technique that we have seen, is being abused by various Internet Service Providers, Router vendors for decades. We have seen in the past, using this technique ISPs, router vendors intercepts HTTP traffic and inject arbitrary data silently into HTTP responses. This is usually done by injecting arbitrary JavaScript code into actual HTTP response body in real time. When the injected JavaScript code reaches client browser it performs various operations such as loading advertisements, information gathering etc.

This paper presents a generic browser exploit detection technique that uses the same Live Network Stream Code Injection technique to reliably catch browser exploits. The detection system can be considered as completely agent less and capable of detecting various techniques, used in modern browser exploitation. Unlike any other Host Based Intrusion Prevention Systems, to be able to generically detect and block browser exploits, no OS API hooking, dll injection or code injection is required in browser process.


Slides & Video Demos:


https://github.com/debasishm89/Brucon0x09

Talk Video:





References:


  • https://brucon0x092017.sched.com/event/BEBO/browser-exploits-grab-them-by-the-collar
  • https://2017.brucon.org/index.php/Debasish_Mandal

Saturday, May 20, 2017

OpenXMolar - A MS OpenXML Format Fuzzing Framework

i) OpenXMolar v 1.0

alt text
OpenXMolar is a Microsoft Open XML file format fuzzing framework, written in Python.

ii) Motivation Behind OpenXMolar

MS OpenXML office files are widely used and the attack surface is huge, due to complexity of the softwares that supports OpenXML format. Office Open XML files are zipped, XML-based file format. I could not find any easy to use OpenXML auditing tools/framework available on the internet which provides software security auditors a easy to use platform using which auditors can write their own test cases and tweak internal structure of Open XML files and run fuzz test (Example : Microsoft Office).

Hence OpenXMolar was developed, using which software security auditors can focus, only on writing test cases for tweaking OpenXML internal (XML and other ) files and the framework takes care of rest of the things like unpacking, packing of OpenXML files, Error handling, etc.

iii) Dependencies

OpenXMolar is written and tested on Python v2.7. OpenXMolar uses following third party libraries

winappdbg / pydbg

Debugger is an immense part of any Fuzzer. Open X-Molar supports two python debugger, one is winappdbg and another is pydbg. Sometimes installing pydbg on windows environment can be painful, and pydbg code base is not well maintained hence winappdbg support added to Open X-Molar. Its recommended that user use winappdbg.

pyautoit

Since we feed random yet valid data into target application during fuzzing, target application reacts in many different ways. During fuzzing the target application may throw different errors through different pop-up windows. To continue the fuzzing process, the fuzzer must handle these pop-up error windows properly. OpenXMolar uses PyAutoIT to suppress different application pop-up windows. PyAutoIt is Python binding for AutoItX3.dll

crash_binning.py

crash_binning is part of sulley framework. crash_binning.py is used only when you've selected pydbg as debugger. crash_binning.py is used to dump crash information. This is only required when you are using pydbg as debugger.

xmltodict

This is not core part of the Open X-Molar. The XML String Mutation module (FileFormatHandlers\xmlHandler.py) was written using xmltodict library.

iv) Architecture:

On a high level, OpenXMolar can be divided into few components.

OpenXMolar.py

This is the core component of this Tool and responsible for doing many important stuffs like the main fuzzing loop.

OfficeFileProcessor.py

This component mostly handles processing of OpenXML document such as packing, unpacking of openxml files, mapping them in memory, converting OpenXML document to python data structures etc.

PopUpKiller.py - PopUp/Error Message Handlers :

This component suppresses/kills unwanted pop-ups appeared during fuzzing.

FileFormatHandlers//

An OpenXML file may contain various files like XML files, Binary files etc. FileFormatHandlers are basically a collection of mutation scripts, responsible for handling different files found inside an OpenXML document and mutate them.

OXDumper.py

OXDumper.py decompresses OpenXML files provided in folder "OpenXMolar\BaseOfficeDocs\OpenXMLFiles" and output a python list of files present in the OpenXML file. OXDumper.py accepts comma separated file extensions. OXDumper.py is useful when you are targeting any specific set of files present in any OpenXML document.

crashSummary.py

crashSummary.py summarizes crashes found during fuzzing process in tabular format. The output of crashSummary.py should look like this:

alt text

v) Configuration File Walk through


The default configuration file 'config.py' is very well commented and explains all of its parameters really well. Please review the default config.py file thoroughly before running the fuzzer to avoid unwanted errors.


vi) Writing your Open XML internal File Mutation Scripts:


As said earlier, an OpenXML file package may contain various files like XML files, Binary files etc. FileFormatHandlers are basically a collection of mutation scripts, responsible for handling different files found inside an OpenXML document and mutate them. Generating effective test cases is the most important step in any fuzz testing process.

The motive behind OpenXMolar was to provide security auditors an easy & flexible platform on which fuzz tester can write their own test cases very easily for OpenXML files. When it comes to effective OpenXML format fuzzing, the main part is how we mutate different files (*.xml, *.bin etc) present inside OpenXML package (zip alike). To give users an idea of how file format handlers are written, two file format handlers are provided with this fuzzer, however they are very dumb in nature and not very effective.

Any file format handler module should be of following structure

# Import whatever you want.
class Handler():# The class name should be always 'Handler'
 def __init__(self):
  pass
 def Fuzzit(self,actual_data_stream): 
  # A function called Fuzzit must be present in Handler class
  # and it should return fuzzed data/xml string/whatever.
  # Note: Data type of actual_data_stream and data_after_mutation should always be same.

  return data_after_mutation

Once your file format handler module is ready you need to place the *.py file in FileFormatHandlers// folder and add the handler entry and associated file extension in config.py file like this :

FILE_FORMAT_HANDLERS = {'xml':'xmlHandler.py',
      'bin':'BinaryHandler.py',
      'rels':'xmlHandler.py',
      'vml':'xmlHandler.py'
      }


vii)Adding More POPUP / Errors Windows Handler

The default PopUpKiller.py file provided with Open X-Molar, is having few most occurred pop up / error windows handler for MS Word, MS Excel & Power Point. Using AutoIT Window Info tool (https://www.autoitscript.com/site/autoit/downloads/) you can add more POPUP / Errors Windows Handlers into 'PopUpKiller.py'. One example is given below.
alt text

So to be able to Handle the error pop up window shown in screen shot, following lines need to be added in : PopUpKiller.py

if "PowerPoint found a problem with content"  in autoit.win_get_text('Microsoft PowerPoint'):
 autoit.control_click("[Class:#32770]", "Button1")

viii)The First Run

This fuzzer is well tested on 32 Bit and 64 Bit Windows Platforms (32 Bit Office Process). All the required libraries are distributed with this fuzzer in 'ExtDepLibs/' folder. Hence if you have installed python v2.7, you are good to go.

To verify everything is at right place, better to run Open X-Molar with Microsoft Default XPS Viewer first time(C:\Windows\System32\xpsrchvw.exe). Place any *.oxps file in '\BaseOfficeDocs\OpenXMLOfficeFiles' and run OpenXMolar.py.

OpenXMolar.py accepts one command line argument which is the configuration file.


C:\Users\John\Desktop\OpenXMolar>python OpenXMolar.py config.py

[Warning] Pydbg was not found. Which is required to run this fuzzer. Install Pydbg First. Ignore if you have winappdbg installed.

   ____                    __   ____  __       _
  / __ \                   \ \ / /  \/  |     | |
 | |  | |_ __   ___ _ __    \ V /| \  / | ___ | | __ _ _ __
 | |  | | '_ \ / _ \ '_ \    > < | |\/| |/ _ \| |/ _` | '__|
 | |__| | |_) |  __/ | | |  / . \| |  | | (_) | | (_| | |
  \____/| .__/ \___|_| |_| /_/ \_\_|  |_|\___/|_|\__,_|_|
        | |
        |_|
        An MS OpenXML File Format Fuzzing Framework.
        Author : Debasish Mandal (twitter.com/debasishm89)

[+] 2017:05:05::23:11:23 Using debugger :  winappdbg
[+] 2017:05:05::23:11:23 POP Up killer Thread started..
[+] 2017:05:05::23:11:24 Loading base files in memory from :  BaseOfficeDocs\UnpackedMSOpenXMLFormatFiles
[+] 2017:05:05::23:11:24 Loading File Format Handler for extension :  xml => xmlHandler.py
[+] 2017:05:05::23:11:24 Loading File Format Handler for extension :  rels => xmlHandler.py
[+] 2017:05:05::23:11:24 Loading File Format Handler Done !!
[+] 2017:05:05::23:11:24 Starting Fuzzing
[+] 2017:05:05::23:11:25 Temp cleaner started...
[+] 2017:05:05::23:11:25 Cleaning Temp Directory...
...
...

ix) Open X-Molar in Action

Here is a very short video on running fuzztest on MS Office Word:

https://www.youtube.com/watch?v=b7n1tuFDl5A

x) Fuzzing Non-OpenXML Applications :

Due to the flexible structure of the fuzzer, this Fuzzer can also be used to fuzz other windows application. You just need do following :

  • In config.py add the target application binary (exe) and extension in APP_LIST of config.py
  • In config.py change OpenXMLFormat to False
  • Write your own File format mutation handler and place it in FileFormatHandlers/ folder
  • Add the newly added FileFormatHandler in FILE_FORMAT_HANDLERS of config.py
  • Provide some base files in folder OtherFileFormats/
  • Add custom error / popup windows handler in PopUpKiller.py using Au3Info tool if required.And you're good to go.

xi) Few More Points about OpenXMolar:

Fuzzing Efficiency: To maximize fuzzing efficiency OpenXMolar doesn't read the provided base files again and from disk. While starting up, it loads all base files in memory and convert them into easy to manage python data structures and mutate them straight from memory.

Auto identification of internal files of OpenXML package : An Open XML file package may contain various files like XML files, Binary files etc. OpenXMolar has capability to identify internal file types and based that chooses mutation script and mutate them. Please refer to the default config.py file (Param : AUTO_IDENTIFY_INTERNAL_FILE_FORAMT) for details.

xii) TODO

Improve Fuzzing Speed
New Feature / Bugs -> https://github.com/debasishm89/OpenXMolar/issues

xiii) Licence

This software is licenced under New BSD License although the following libraries are included with Open X-Molar and are licensed separately.

xiv) Source 


The source code is available here : https://github.com/debasishm89/OpenXMolar