Re: GSoC IPv6 Machine learning

["Mathias Morbitzer" <m.morbitzer () runbox com>]

Hi, 

Last summer, we have completely redesigned the IPv6 OS detection system to be able to cope with the new challenges of 
IPv6, minimize the manual effort needed to add detection of new operating systems and to improve behavior when 
confronted with unknown operating systems.  

This current system uses logistic regression combined with an one-vs-all classification. To be able to cope with 
missing values, they are imputed with the multiple imputation by chained equations (MICE) approach.

One of the things we would like to do is to analyze if new probes can be added, modified or removed to increase 
efficiency of the IPv6 OS detection engine.  Also for the set of extracted features research needs to be done with 
regards to features that can be added or removed. Further ideas for improvement of the OS detection engine are also 
welcome.

Also, given the structure of the data, there might be other, more suited approaches to perform classification, such as 
a random forest classifier. Therefore another idea would be to design and implement an alternative system for OS 
detection within Nmap, and to see if the new system is able to perform better than the one currently in use.

I hope this answers your questions. 

Regards,
Mathias

On Fri, 18 Mar 2016 15:27:05 -0700 (MST), tamimcsedu19 <tamim.tamim1382 () gmail com> wrote:

> I was wondering what would be the goal of the project as. Should we use svm ,
> logistic regression , decision trees and other techniques and compare the
> results of which works best for the task ?
>
>
>
> --
> View this message in context: http://nmap-dev.996309.n3.nabble.com/GSoC-IPv6-Machine-learning-tp27457.html
> Sent from the Nmap - Dev mailing list archive at Nabble.com.
> _______________________________________________
> Sent through the dev mailing list
> https://nmap.org/mailman/listinfo/dev
> Archived at http://seclists.org/nmap-dev/


_______________________________________________
Sent through the dev mailing list
https://nmap.org/mailman/listinfo/dev
Archived at http://seclists.org/nmap-dev/