====== Transmitter Identification ======

==== Files ====
Mono receiver: [[http://xp.cortexlab.fr/storage/monoRxSet_012019.tgz|Dataset]]

The setup used to generate this data (GNURadio blocks and flowgraphs, FIT/CorteXlab scenarios and example usage) can be found in [[https://github.com/Inria-Maracas/gr-txid|this GitHub repository]]


Licence: [[https://creativecommons.org/licenses/by/4.0/|Creative Commons Attribution 4.0]]

==== Background ====
Hardware imperfections in RF transmitters introduce features that can be used to identify a specific transmitter among others.
Currently, header size sometimes outweigh the payload size in IoT type small packets. Furthermore, headers are currently the only barrier against transmitter identification errors and transmitter impersonation on edge devices that don’t have the resources to use cryptographic protocols.
Therefore, a system able to identify a transmitter based on intrinsic hardware features could help reduce packet sizes and/or improve security.


==== Content description ====
This is a compilation of the results of experiments run on the FIT/CorteXlab platform under different scenarios as described in the paper [[https://arxiv.org/abs/1905.07923|Transmitter Classification With Supervised
Deep Learning]].

The archive contains a folder for each individual experiment, named after the time it was run.
Inside of each is a small //Readme// file with a summary of the experiment parameters.

The data is stored inside of .bin files.
The ones starting with ''im'' store the imaginary part of the complex samples
and the ones starting with ''re'' store the real part.
The last number corresponds to the index of the transmitter that sent the samples.   
It gives something like this:
''<im/re>_00_<tx_id>.bin''

Inside of these, samples are in float32 with 600 of them making a packet.

The git repository contains an example of how to load the data in python and then serialize it with pickle for later use.