WAPFinder evolves
Hi all,
one of my encog based projects is close to release.
The application takes the log files from kismet wireless ( http://www.kismetwireless.net/ ) and analyzes the data to estimate the location of the access point.
As mathematical approaches in complex environments did not succeed, I gave encog a try ... finally it seems to work, but I need a few other people to test it.
You can find some screenshots here: http://scacs.no-ip.org:83/gallery2/v/wapfinder/
Requirements for tests:
* kismet xml and gps files from a scan (you should create them by your own)
* Windows PC with .NET
Requirements for running a Kismet scan:
* Linux device with attached WLAN card and GPS reciever.
* After the run, the kismet log files must be copied to your Windows PC.
I use a backtrack live CD ( http://www.remote-exploit.org ) and a simple usb GPS reciever for this.
The final application will be closed source, but free for private use.
If anybody is interested in running some test calculations, please contact me!
RSS
YouTube
Twitter
Facebook
github
LinkedIn
That looks very interesting, and nice screen shots.
When you have it complete, send me a link, and I will add it to the other projects that use Encog.
Of course, as encog is the best AI I found!
btw: ART3 would be nice here to analyze the output :-)
Hendrik
-- short update --
The calculation description:
The calculation is based on the values of latitude, longitude and signal.
- In the first step, the boundaries of the evaluation fields are calculated based on the selected BSSID.
- The next step sets the ZERO points. ZERO points are physically approximated points, where no signal will be detectable.
- The third step is optional and tries to optimize the neural network for the given points.
- As fourth step, the neural network is trained multiple times.
- The last step calculates the weighted mean of the values and displays it.
ART 2 and 3 are defiantly on the radar for future releases.
I know; this was only a broad hint from my side :-)