search

doyensec / inql

InQL is a robust, open-source Burp Suite extension for advanced GraphQL testing, offering intuitive vulnerability detection, customizable scans, and seamless Burp integration.
https://doyensec.com/
Apache License 2.0
1.54k stars 158 forks source link

[Improper parsing]: Hackerone #129

Open abhinavsecond opened 1 year ago

abhinavsecond commented 1 year ago

InQL version

BApp Store

GraphQL API

https://hackerone.com/graphql

GraphQL specification version

No response

What isn't working?

I have waited more that 30+ min to show the results of HackerOne Graphql schemas but it is not showing any result Note:- You need to be authenticated to query graphql on hackerone image It is stuck a this point

abhinavsecond commented 1 year ago

Well after 10 more mins the result finally appeared but blank image

execveat commented 1 year ago

Thank you for the report! The Activity query alone takes 2 minutes to get parsed on my machine. I filed the issue with GQLSpection, will take a look at optimizations we could add here: https://github.com/doyensec/GQLSpection/issues/32

The next InQL version contains syntax highlighting which causes some additional drop in performance, so we definitely need to address it.

abhinavsecond commented 1 year ago

Thanks for looking into it🙂 I also noticed that while using this, it's using a lot of resources in Burp Suite, up to 7.2 GB of RAM. image I was running it on my main computer, which has plenty of RAM, so it didn't crash Burp. If it was on a virtual machine (VM), I think it might have caused the VM to crash. I believe you should investigate this further because even though it's a great tool, if it crashes a user's Burp Suite or VM, they won't be able to use it effectively.

Thanks Abhinav

execveat commented 1 year ago

So, this turned out to be way more complicated than I initially thought. I'm planning to address the performance issue by reducing recursion to an iterative loop and generating queries on demand (so, only when you click one, not ahead of time). Initially this was planned for a future release, but it looks like a release breaking issue now.

In case of H1 specifically, some queries right now are more than 1GB in size which explains the observed RAM usage. Obviously, this size makes them completely impractical to work with. So we could just add a sanity check that detects these huge queries and refuses to work with them, but obviously these problematic queries might very well be the most interesting ones, so in my view you should be able to work with them using Burp & InQL (on a beefy machine).

So, it's a complicated problem that will likely need to be addressed both from performance tuning perspective (everything's single core right now...) as well as from the UI/UX to intelligently disable heavyweight features such as syntax highlighting and inline comments for larger queries. Stay tuned.

execveat commented 1 year ago

I replaced the recursion with a loop and the parsing time for the whole h1 schema (on my machine) now is under 2 minutes. I will try to reduce this loading time further by generating queries on demand (upon the click) instead of ahead-of-time, as well as introducing parallelization.

You can try the new version, but it's not release ready yet. Clicking larger queries (like activity) completely freezes Burp, presumably due to expensive highlighting routine. I'll try to introduce auto-disabling highlighting for larger queries as a stop-gap measure, followed by performance optimization.

marz-hunter commented 7 months ago

I also experienced this problem. I load from introspection (json file) takes -+ 45 minutes. but when clicking to view it can't be seen (I'm using the version from the baap store 10 oct 2023)