Parsing embedded JSON

[geekscrapy]

Hi,

I have the following JSON format:

{"dst_host": "159.65.224.130", "dst_port": 23, "honeycred": false, "local_time": "2019-01-23 11:57:11.834296", "logdata": {"PASSWORD": "1111", "USERNAME": "root"}, "logtype": 6001, "node_id": "opencanary-1", "src_host": "41.139.253.2", "src_port": 36653}

How would I get agrind to also parse the logdata field?

TIA!

[rcoh]

Nested JSON isn't super well supported right now, but try: a agrind '* | json | json from logdata:

echo '{"dst_host": "159.65.224.130", "dst_port": 23, "honeycred": false, "local_time": "2019-01-23 11:57:11.834296", "logdata": {"PASSWORD": "1111", "USERNAME": "root"}, "logtype": 6001, "node_id": "opencanary-1", "src_host": "41.139.253.2", "src_port": 36653}' 
| agrind '* | json | json from logdata | fields PASSWORD, USERNAME'
[PASSWORD=1111]            [USERNAME=root]

You can use the fields operator to drop the other fields if you want.

[geekscrapy]

Awesome, thanks!

Appreciate it's not trivial, however the main usecase for me would be to read JSON converted EVTX (windows logs) where there may (or may not be) multiple levels of json... Would be killer if that would be supported

[rcoh]

Yeah, makes sense. I assume you don't know the keys ahead of time? What would the ideal workflow be for you? Something like:

* | json | count by logdata.user?

[rcoh]

Keeping the issue open to discuss longer term improvements

[geekscrapy]

Yea, the keys relate to the event id (and there are thousands of event ids....).

Btw you probably know, but that example JSON I gave earlier is a random log I had floating around, it's not a Windows EVTX. I could provide you one if you wanted (or you can just copy one from a Windows machine).

The original format of EVTX is binary XML, which is usually extracted using the following python library. This XML is usually then converted to a python dict, then to JSON format. So it's a bit of a pain, but if you could shortcut the conversion process, that'd be a massive win. There are very few tools that allow analysis of EVTX on Linux. But maybe there is a good reason for this and I'm missing it 😂

https://github.com/williballenthin/python-evtx/issues/47

[rcoh]

yeah I'm probably not going to at specific EVTX support...I could add support to redirect the pipeline to another program though

On Sun, Jan 27, 2019 at 2:56 PM molley notifications@github.com wrote:

Yea, the keys relate to the event id (and there are thousands of event ids....).

Btw you probably know, but that example JSON I gave earlier is a random log I had floating around, it's not a Windows EVTX. I could provide you one if you wanted (or you can just copy one from a Windows machine).

The original format of EVTX is binary XML, which is usually extracted using the following python library. This XML is usually then converted to a python dict, then to JSON format. So it's a bit of a pain, but if you could shortcut the conversion process, that'd be a massive win. There are very few tools that allow analysis of EVTX on Linux. But maybe there is a good reason for this and I'm missing it 😂

williballenthin/python-evtx#47 https://github.com/williballenthin/python-evtx/issues/47

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/rcoh/angle-grinder/issues/64#issuecomment-457962271, or mute the thread https://github.com/notifications/unsubscribe-auth/AAeFZyzqYnbKCJM24YNXGBfhvTZhpYiWks5vHi6hgaJpZM4aUn6l .

[geekscrapy]

No that's fine, I was more hoping for either nested XML or JSON. We only really need one of those and it'll be covered.

Adding an option to redirect seems a little more complicated than it needs to be I think. cat works fine 😀

[rcoh]

You might also try JQ to do whatever complex JSON munging you want to do before piping the result to angle-grinder

On Sun, Jan 27, 2019 at 10:58 PM molley notifications@github.com wrote:

No that's fine, I was more hoping for either nested XML or JSON. We only really need one of those and it'll be covered.

Adding an option to redirect seems a little more complicated than it needs to be I think. cat works fine 😀

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/rcoh/angle-grinder/issues/64#issuecomment-458017664, or mute the thread https://github.com/notifications/unsubscribe-auth/AAeFZ5SUVHoGVeD5XGVNnU06dHQo-Qbdks5vHp-agaJpZM4aUn6l .

[rcoh]

I think a good solution for this could be splatting JSON objects into 1-row-per KV like https://github.com/tomnomnom/gron -- @geekscrapy I'm curious if that would create a usable output for EVTX JSON

[geekscrapy]

Hey! This may work (disclaimer: I've not played with gron, but it sounds like it should work)

[rcoh]

With #73 adding a splat operator should be all that's required to get decent support for arbitrary nested structures.

[Arch-vile]

I tried running the example you gave earlier (dropped some fields here for clarity): echo '{"dst_host": "159.65.224.130", "logdata": {"PASSWORD": "1111", "USERNAME": "root"} }' | agrind '* | json | json from logdata | fields PASSWORD, USERNAME'

But end up with: error: Expected string, found other

agrind --version
ag 0.18.0

[rcoh]

ah, that was probably written before proper nested field support was added.

You can now refer to logdata.PASSWORD and logdata.USERNAME directly. If you want to restructure things, you can do it like this: * | json | logdata.PASSWORD as password | logdata.USERNAME as username | fields username, password:

echo '{"dst_host": "159.65.224.130", "logdata": {"PASSWORD": "1111", "USERNAME": "root"} }' | agrind '* | json | logdata.PASSWORD as password | logdata.USERNAME as username | fields username, password'
[password=1111]            [username=root]