The Great Escape TryHackMe Writeup

5 minute read


The Great Escape is a medium rated Linux based room on tryhackme by hydragyrum. SSRF along with code injection was used to get a root shell on a docker container. Using Port knocking sequence, TCP port for Docker container was opened which was used to get a root shell on the box.

Port Scan

Full Port Scan

[email protected]:~/Documents/tryhackme/thegreateescape$ nmap -p- -min-rate 10000 -v -oN nmap/all-ports
Nmap scan report for
Host is up (0.32s latency).
Not shown: 65366 closed ports, 167 filtered ports
22/tcp open  ssh
80/tcp open  http

Read data files from: /usr/bin/../share/nmap
# Nmap done at Mon Feb 15 01:21:34 2021 -- 1 IP address (1 host up) scanned in 31.43 seconds

Only two ports are open. SSH is running on port 22 and HTTP service is running on port 80.

Detail Scan

[email protected]:~/Documents/tryhackme/thegreateescape$ nmap -p22,80 -sC -sV -oN nmap/detail
Starting Nmap 7.80 ( ) at 2021-02-15 21:16 +0545
Nmap scan report for
Host is up (0.31s latency).

22/tcp open  ssh?
| fingerprint-strings: 
|   GenericLines: 
|_    im4+ ^1a}wVQbdyQ/{VY!;
|_ssh-hostkey: ERROR: Script execution failed (use -d to debug)
80/tcp open  http    nginx 1.19.6
| http-robots.txt: 3 disallowed entries 
|_/api/ /exif-util /*.bak.txt$
|_http-server-header: nginx/1.19.6
|_http-title: docker-escape-nuxt
1 service unrecognized despite returning data. If you know the service/version, please submit the following fingerprint at :

Service detection performed. Please report any incorrect results at .
Nmap done: 1 IP address (1 host up) scanned in 195.25 seconds

Let us start the enumeration with HTTP service on port 80.

HTTP service on Port 80


Clicking on /courses, /admin redirected me to /login and the signup seems to be disabled.


Normally my next step would be directory and files bruteforcing, but this webapp has a firewall in place which has a rate limiting.

Entries on robots.txt

[email protected]:~/Documents/tryhackme/thegreateescape$ curl
User-agent: *
Allow: /
Disallow: /api/
# Disallow: /exif-util
Disallow: /*.bak.txt$

We can see that there are three entries on robots.txt

Checking /exif-util

3 Looking at the title of the webpage Exif Utlis, it looks like we can upload a file and this app will display the metadata of that file. Also the file can be included by an URL.

Uploading a test image

4 I uploaded a PNG image and the app gave the information about the image. We can see that the app is making a request to /api/exif and the file name of the uploaded file is also changed.

Including image via URL

I have hosted files using python server from my device. 5 This time we made a GET request to /api/exif with a parameter called url.

Checking if the URL parameter is vulnerable to SSRF

From the Nmap scan,we saw that port 22 and 80 are open.

Checking whether we get a response from port 80

6 I did not get any response either from port 80 or 22. As I was manually poking, I got reponse from port 8080.

Port 8080

7 Since we know that our host has port 22 and 80 open, I thought that this webapp must be running inside a docker container.

Checking for port 80 on docker’s interface

8 And my assumption was correct. Using this we can scan the whole network for open ports. Since there was rate limiting implemented, I did not bother to go down that path.

I checked for the other protocols like file,zip,gopher, but all of them gave some sort of error.

File Bruteforcing

There is a interesting entry on robots.txt. ie *.bak.txt. Since there is rate limiting on the webserver, instead of using the usual wordlists for bruteforcing, I manually created a small custom wordlist with the words present on the webserver.

Content of the wordlist


Using wfuzz with a single thread


Content of exif-util.bak.txt

10 We get a hostname. Looks like it is a api’s backup docker container used by the dev team. Since this is a backup and a developement container, it might contain unpatched vulnerabilities.

Checking for SSRF on backup container

11 For the empty value of the url parameter, we get a error with curl on it which means that the value of the parameter is directly passed to the curl.


url = $_GET['url'];
system('curl ' . 'url');


If this is the case, we can get code execution.

Trying for command injection

12 And we successfully executed the command. We are root, but just on the docker container.

I tried to get a reverse shell but was unsuccessful. It looks like the all the outgoing traffic is blocked by the firewall. So, I manually started going through the container.

On /root


Content of dev-note.txt

14 I tried to login with hydra:fluffybunnies123 using SSH, but was unsuccessful.

Enumerating the git repo

15 We can see that are three commits. So, lets check those out.

Checking the commits

16 We get a port knocking sequence for opening docker TCP port and a flag too.

Performing Port Knocking Sequence


Full Port Scan

Let’s run nmap to check for any new open ports.

[email protected]:~/Documents/tryhackme/thegreateescape$ nmap -p- --min-rate 10000 -v
Starting Nmap 7.80 ( ) at 2021-02-15 22:42 +0545
Nmap scan report for
Host is up (0.31s latency).
Not shown: 65528 closed ports
22/tcp    open     ssh
80/tcp    open     http
2375/tcp  open     docker

Read data files from: /usr/bin/../share/nmap
Nmap done: 1 IP address (1 host up) scanned in 28.72 seconds

And we get a new port ie 2375.

I searched and found this amazing article which explains that we can abuse docker API to get a root shell on the box.

Listing docker images


We can successfully list the docker images using the API.

Getting a root shell


Here I have created a container from frontend image on an interactive mode executing sh binary. The root file system of the host will be mounted on the /mnt directory of the container and the root of the container is changed to /mnt.

Reading web flag



Reading root flag


Having trouble with the writeup and need help, do not hesitate to ping me on twitter.