GSM Radio Frequency Fingerprinting

Citation Author(s):
Gianmarco
Baldini
Submitted by:
Gianmarco Baldini
Last updated:
Tue, 01/03/2023 - 09:41
DOI:
10.21227/c7sa-wq24
License:
0
0 ratings - Please login to submit your rating.

Abstract 

This is a data set for Radio Frequency fingerprinting, which is a kind of identification of wireless devices based on their intrinsic physical features. The data set is composed by GSM bursts collected from 12 GSM mobile phones while transmitting.  The samples have been collected using a Software Defined Radio with a sample rate at 20 MS/s. The content information has been removed from the bursts to remove the risk of bias due to content. The data set is in MATLAB format.

Instructions: 

This is a data set useful for RF fingerprinting research. It is a collection of signals (I/Q) collected from 12 different mobile phones of 4 different models (HTC One, Sony, Samung S5 and Nokia). The GSM standard is used. There are three phones for each model. A set of 1000 bursts have been collected. The data is represented as a three dimensional array in MATLAB format. The first dimension is the identifier of the burst, the second dimension is the burst itself in complex number format (I/Q) and the third dimension is the identifier of the mobile phone. In each burst, the data field has been removed and only the remaining fields (transients, preambles and so on) are left. The system is configured in such a way that the preambles and the other fields are invariant across the phones.

In the MATLAB shell, the command load can be used to load the contents of the file. There is only the array as a data structure in the file:

load('GSMData');

The data set can be used directly for machine learning classification to implement RF identification based on the RF fingerprints because the data is synchronized and normalized using the RMS function.