From . in regex to SSRF - part 2

Posted on Jan 14, 2020

In this story I will continue my journey to exploit Server Side Request Forgery (SSRF). Part 1 is available here.

What is state of the game up to now? I have found service that is vulnerable to SSRF by executing the REST call and by passing domain name check:

I have registered domain name: and add for it CNAME record to my EC2 server.

But wait a second. This is not yet “SSRF”. For now I could call specific domain, but didn’t prove ability to scan internal ports or execute requests to internal services.

First of all I checked what headers are sent to my server:

nc -l -n -p 80

What it interesting here is User-Agent. Clearly indicating NodeJS and node-fetch library.


For this part I needed http server to host some files and do redirects. Here is my script based on python http.server:

This script is doing two things:

  • if request is starting with query parameter /?r= it is taking value of this parameter and put it in Location header, returning code 302
  • else it is loading file from disk and always returning image/svg+xml with code 200. This is important! Always return expected Content-Type to check if parser can handle different types.

This code is PoC so don’t expect much out of it. It’s for testing purpose only. Don't use it on production.

You can check more of my hacking resources in my public repo.

Mind map

Here is mind map of ideas that I had during exploitation:

SVG payload

Hosting some svg files was my first shot. I have took some from PayloadsAllTheThings repository, but sadly none was working.

In most cases I got 502 Bad Gateway and in some payloads were just ignored. The other thing I noticed is that I was able request png or jpg file and it was parsed. I have gut feeling that service was using some kind of NodeJS library, not ImageMagick. Maybe I did miss something?

HTTP Redirect

I have implemented in my possibility to redirect with new Location header. This created variety of options to exploit.

Lets start with ports scanning:

Redirect part is url encoded to not break service with ? or /. I have taken this to Burp Intruder and scan all ports. But if you are using free version of Burp this can take a very long time. So it’s better to use ffuf:

for i in {1..65535}; do echo $i; done > all_ports.txt;
ffuf -mc all -ac -w all_ports.txt -u ''

Same way I was able to scan for internal services. For list of domains I used SecLists:

ffuf -mc all -ac -w /usr/share/wordlists/SecLists/Discovery/Web-Content/common.txt -u ''

This will try to make a call to subdomains of but from perspective of In my case it found redis instance on

Next step for me was access cloud resources, e.g. metadata services on

I was able to positively connect to So it means that service was running on AWS.


I finally got blind SSRF using redirect 302 code and Location header! That’s gave me possibility to penetrate internal network. I was disappointed that no svg payload was working. For part 3 I left “HTTP Parser Abuse” - I will try to force node-fetch to change protocol and talk with services like redis.

Thanks for reading! You can follow me on Twitter.