From . in regex to SSRF - part 3
This part is focused on abusing Node.js and node-fetch library. I will try to “talk” with Redis service using CRLF injection in http parser.
For convenience Redis service will be simulated by
nc -vvlp 6379.
Test environment from my Kali 2020.1b:
- Node.js version 10.19.0
- node-fetch version 2.6.0
Lets start with PayloadsAllTheThings. It contains couple of CRLF Injection payloads. I will loop over them and check result in second console:
None success here. All payloads failed 🙁
－＊Set-Cookie:injection－＊ (Unicode U+FF0D U+FF0A) http://0\r\n SET foo 0 60 5\r\n :6379/ https://0\r\nSET foo 0 60 5\r\n:6379/
Still no success here. I seams that this version of Node.js is not vulnerable for CRLF attacks.
Let’s try harder and dig dipper into node-fetch, maybe something interesting will be in code 😃
Investigation of node-fetch code
What am I trying to achieve here? I have in mind two types of possible errors:
- Url parsing
- Handling url input as object not as string
Let’s see what I will find.
Debug of Node.js code is quite nice with Visual Studio Code:
Problem number one is not existing as node-fetch is using standard Node.js
Url.parse for input. There are not doing much fancy stuff with it.
For second problem I needed to do more investigation.
First of all I will explain why I’m interested in processing object instead of string. In many dynamic languages you can make valid request like this:
This leads to created object instead of string. Could be quite handy for some scenarios. Especially if developers didn’t predict it 😃 See below example of parsing such url in Node.js Express framework.
In node-fetch I have found one possible attacking vector:
It look like possible to use object instead of string for input parameter. This
input.method could change method type in some specific conditions. After spending some time in debugger it turn out as dead end.
I didn’t manage to escalate blind SSRF to anything more. I have spent couple of days trying different approaches. Nevertheless after submitting report I was awarded with 400$ and bug was marked as medium.
Thanks for reading! You can follow me on Twitter.