Skip to content

Neural Network for identifying people from channel state information provided by a Nexus 5 phone via Nexmon Firmware patch, via beaconing packets of a router

Notifications You must be signed in to change notification settings

aldesilv/NexmonNeuralNetwork

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Identifying People with Wifi

This project was able to identify the 4 individuals and the empty room with 86 percent accuracy using only the channel state information of a Nexus 5 phone that had the Nexmon Firmware patch. A router was set in a specific configuration as to not hop channels (for ease of use) and then it sent out beaconing packets. Individuals walked one at a time past the cell phone (Nexus 5).

The data from the Nexus 5 was a .pcap file (see the pcap-data folder). This file was put into the csi_parser.py script to convert it to a .csv. In the original experiment the csv underwent normalization and was changed into an image in Matlab. However, Tensorflow/Keras can handle .csv files directly and can do normalization itself.

Findings from Spring 2018

Our findings seem to indicate that the Neural Network identifies humans based on Gait.

Gait

Here is the Neural Network that was most effective.

Model

More information in the .pdf inside the paper folder

Information about Spring 2018 Data

Table and Patterns

Each person walked for 30-40 minutes over all (After the data was gathered and fed into the NN this is clearly more then nessisary for 1 room of data, instead of sending in 30 seconds per frame it is likely the same results can be done with 5-8 seconds) with 5-7 minutes per method (multiple walking paths past the router cell phone for the fast/slow walks).

About

Neural Network for identifying people from channel state information provided by a Nexus 5 phone via Nexmon Firmware patch, via beaconing packets of a router

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 80.9%
  • Python 15.9%
  • Shell 3.2%