Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Machine learning #32

Open
void-elf opened this issue Feb 1, 2019 · 7 comments
Open

Machine learning #32

void-elf opened this issue Feb 1, 2019 · 7 comments
Labels
enhancement New feature or request research Watchdog Related to the watchdog code

Comments

@void-elf
Copy link
Collaborator

void-elf commented Feb 1, 2019

Some initial steps:

  • @ynasser to contact ML researchers we collaborated with in the past
  • start saving binary blog pcaps

Also, UW researchers suggested clustering and unsupervised learning, and then an iterative labeling approach.

@void-elf void-elf added enhancement New feature or request research labels Feb 1, 2019
@cooperq cooperq added the Watchdog Related to the watchdog code label Feb 15, 2019
@cooperq cooperq added this to the Scanning Enhancements milestone Jul 11, 2019
@marcfielding1
Copy link

Do you have training data available for this? Ie data with verified bogus cell sites? We can donate some compute to this if you like.

@cooperq
Copy link
Collaborator

cooperq commented Aug 14, 2020

Hoping to gather some training data with crocodile hunter. We had considered doing an unsupervised algorithm but my (very limited) understanding is that this would require a TON of training data

@marcfielding1
Copy link

marcfielding1 commented Aug 16, 2020

Just copying this over from #83 to keep the train of thought in a single thread.

Also if you're recording the data gathered by crocodile hunter we can start experimenting with (this thread) (machine learning) to provide a probability score really you need as much data as possible on each mast then labelled samples of fake
ones - which in itself isn't a hard thing to do.

If you already have a ton of mast data for anywhere let me know I can try some basic unsupervised learning methods and see how it does.

Out of interest(forgive my ignorance) but the vulns you discussed in 4g in terms of handshake, as I understand the handshake is where things get funky, what if you recorded the handshake across as many masts as possible and you could weed out the ones the did odd stuff(like downgrading to 2g)?

@cooperq
Copy link
Collaborator

cooperq commented Aug 17, 2020

yea if we could record the handshake that would be great. We can't do that from CH because that would require transmitting which the SDR isn't licenesed for. But if we could record the handshake from a 4G modem that would be ideal.

@marcfielding1
Copy link

Ahhh yeah I remember now I was just about to look back on your video, hrmm I've got loads of them knocking about but I've no idea how I'd record it at the moment, I'll do some digging, really it'd need to absorb masts from CH scans and then be instructed to connect to each one and record the handshake somehow I guess.

@marcfielding1
Copy link

Looks like something like this might work https://desowin.org/usbpcap/ - I'll keep researching.

@tonyayoub23
Copy link

machine learning what is this !! https://appslite-ar.com/

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request research Watchdog Related to the watchdog code
Projects
None yet
Development

No branches or pull requests

4 participants