Methods, architectures, apparatuses and systems directed to data augmentation of radio frequency (RF) data for improved RF fingerprinting
Procedures, methods, architectures, apparatuses, systems, devices, and computer program products directed to data augmentation of radio frequency (RF) data for improved RF fingerprinting are provided. Among the methods is method that may include any of obtaining one or more samples by sampling a radio frequency (RF) signal received at a receiver from a transmitter; determining one or more channel characteristics of a channel between the receiver and the transmitter; and performing RF fingerprinting based at least in part on (i) inputting the samples and the channel characteristics as inputs to a neural network formed using a trained neural network model, and (ii) obtaining a predicted value output from the neural network.
1 . A method comprising:
obtaining one or more samples by sampling a radio frequency (RF) signal received at a receiver from a transmitter;
obtaining one or more channel characteristics of a channel between the receiver and the transmitter; and
performing RF fingerprinting based at least in part on:
inputting the one or more samples and the one or more channel characteristics as inputs to a neural network; and
obtaining a predicted value output from the neural network.
2 . The method of claim 1 , wherein one or more statistical channel descriptions are input to the neural network in connection with performing RF fingerprinting.
3 . The method of claim 1 , comprising:
determining that an identifier of the transmitter is spoofed or not spoofed based on the predicted value and one or more criteria.
4 . The method of claim 3 , wherein one or more statistical channel descriptions are input to the neural network in connection with performing spoof detection.
5 . The method of claim 1 , wherein the neural network is formed using a trained neural network model.
6 . The method of claim 5 , comprising:
determining a statistical description of the channel based at least in part on the one or more channel characteristics; and
obtaining the trained neural network model based on the statistical description of the channel.
7 . The method of claim 6 , wherein obtaining the trained neural network model comprises:
selecting the trained neural network model from a repository, wherein the trained neural network model is selected based on the statistical description.
8 . The method of claim 6 , wherein obtaining the trained neural network model comprises:
determining the trained neural network model based on the statistical description.
9 . The method of claim 8 , wherein the one or more samples are one or more first samples, and wherein determining the trained neural network model comprises:
augmenting one or more second samples of one or more RF signals based on one or more signal impairments; and
training an untrained neural network or a partially trained neural network using the augmented one or more second samples.
10 . The method of claim 9 , wherein the one or more signal impairments comprise any of:
one or more signal impairments other than those associated or attributed to a transmitter;
one or more signal impairments that occur after transmission; and
one or more propagation characteristics typical of a channel between a transmitter and a receiver.
11 . The method of claim 9 , wherein augmenting one or more second samples of one or more RF signals comprises:
augmenting the one or more second samples using one or more transforms that mimic the one or more signal impairments.
12 . The method of claim 11 , comprising:
generating the one or more transforms based on the one or more signal impairments, and wherein augmenting the one or more second samples comprises:
applying the one or more transforms to the one or more second samples.
13 . The method of claim 12 , wherein at least one of the one or more transforms mimics a realistic channel impulse response.
14 . The method of claim 13 , wherein the at least one of the one or more transforms comprises a finite impulse response filter.
15 . The method of claim 12 , wherein at least one of the one or more transforms mimics a realistic channel frequency spectrum.
16 . The method of claim 15 , wherein the at least one of the one or more transforms comprises a frequency domain filter.
17 . The method of claim 16 , wherein at least one of the one or more transforms mimics a realistic channel frequency spectrum, and wherein applying the one or more transforms to the one or more second samples comprises: performing element-wise multiplication of the one or more second samples with filter coefficients of the frequency domain filter.
18 . The method of claim 9 , comprising:
saving the trained neural network as a neural network model.
19 . The method of claim 9 , comprising:
labeling the trained neural network.
20 . The method of claim 19 , wherein labeling the trained neural network comprises:
labeling the trained neural network based on any of a characteristic of a network in which the transmitter is deployed and a characteristic of an environment in which the transmitter is deployed.
21 . The method of claim 19 , comprising:
saving the labeled, trained neural network as a neural network model.
22 . The method of claim 8 , wherein the one or more samples are one or more first samples, and wherein determining the trained neural network model comprises:
augmenting one or more second samples of one or more RF signals based on one or more propagation characteristics typical of a channel between a transmitter and a receiver; and
training an untrained neural network or a partially trained neural network using the augmented one or more second samples.
23 . The method of claim 22 , wherein augmenting one or more second samples of one or more RF signals comprises:
augmenting the one or more second samples using one or more transforms that mimic the one or more propagation characteristics.
24 . The method of claim 23 , comprising:
generating the one or more transforms based on the one or more propagation characteristics, and wherein augmenting the one or more second samples comprises:
applying the one or more transforms to the one or more second samples.
25 . The method of claim 1 , wherein the one or more channel characteristics comprise any of an estimated channel input response and an accumulation of estimated channel input response.
26 . An apparatus comprising circuitry, including any one or more of a processor and memory, the circuitry configured to:
obtain one or more samples by sampling a radio frequency (RF) signal received at a receiver from a transmitter;
obtain one or more channel characteristics of a channel between the receiver and the transmitter; and
perform RF fingerprinting based at least in part on:
inputting the one or more samples and the one or more channel characteristics as inputs to a neural network; and
obtaining a predicted value output from the neural network.
27 . The apparatus of claim 26 , wherein one or more statistical channel descriptions are input to the neural network in connection with performing RF fingerprinting.
28 . The apparatus of claim 26 , wherein the circuitry is configured to:
determine that an identifier of the transmitter is spoofed or not spoofed based on the predicted value and one or more criteria.
29 . The apparatus of claim 28 , wherein one or more statistical channel descriptions are input to the neural network in connection with performing spoof detection.
30 . The apparatus of claim 26 , wherein the neural network is formed using a trained neural network model.
31 . The apparatus of claim 30 , wherein the circuitry is configured to:
determine a statistical description of the channel based at least in part on the one or more channel characteristics; and
obtain the trained neural network model based on the statistical description of the channel.
32 . The apparatus of claim 31 , wherein, to obtain the trained neural network model, the circuitry is configured to:
select the trained neural network model from a repository, wherein the trained neural network model is selected based on the statistical description.
33 . The apparatus of claim 31 , wherein, to obtain the trained neural network model, the circuitry is configured to:
determine the trained neural network model based on the statistical description.
34 . The apparatus of claim 33 , wherein the one or more samples are one or more first samples, and wherein, to determine the trained neural network model, the circuitry is configured to:
augment one or more second samples of one or more RF signals based on one or more signal impairments; and
train an untrained neural network or a partially trained neural network using the augmented one or more second samples.
35 . The apparatus of claim 34 , wherein the one or more signal impairments comprise any of:
one or more signal impairments other than those associated or attributed to a transmitter;
one or more signal impairments that occur after transmission; and
one or more propagation characteristics typical of a channel between a transmitter and a receiver.
36 . The apparatus of claim 34 , wherein the circuitry is configured to:
augment the one or more second samples using one or more transforms that mimic the one or more signal impairments.
37 . The apparatus of claim 36 , wherein the circuitry is configured to:
generate the one or more transforms based on the one or more signal impairments; and
apply the one or more transforms to the one or more second samples.
38 . The apparatus of claim 37 , wherein at least one of the one or more transforms mimics a realistic channel impulse response.
39 . The apparatus of claim 38 , wherein the at least one of the one or more transforms comprises a finite impulse response filter.
40 . The apparatus of claim 37 , wherein at least one of the one or more transforms mimics a realistic channel frequency spectrum.
41 . The apparatus of claim 40 , wherein the at least one of the one or more transforms comprises a frequency domain filter.
42 . The apparatus of claim 41 , wherein at least one of the one or more transforms mimics a realistic channel frequency spectrum, and wherein the circuitry is configured to perform element-wise multiplication of the one or more second samples with filter coefficients of the frequency domain filter.
43 . The apparatus of claim 34 , wherein the circuitry is configured to:
save the trained neural network as a neural network model.
44 . The apparatus of claim 34 , wherein the circuitry is configured to:
label the trained neural network.
45 . The apparatus of claim 44 , wherein the circuitry is configured to:
label the trained neural network based on any of a characteristic of a network in which the transmitter is deployed and a characteristic of an environment in which the transmitter is deployed.
46 . The apparatus of claim 44 , wherein the circuitry is configured to:
save the labeled, trained neural network as a neural network model.
47 . The apparatus of claim 33 , wherein the one or more samples are one or more first samples, and wherein, to determine the trained neural network model, the circuitry is configured to:
augment one or more second samples of one or more RF signals based on one or more propagation characteristics typical of a channel between a transmitter and a receiver; and
train an untrained neural network or a partially trained neural network using the augmented one or more second samples.
48 . The apparatus of claim 47 , wherein the circuitry is configured to:
augment the one or more second samples using one or more transforms that mimic the one or more propagation characteristics.
49 . The apparatus of claim 48 , wherein the circuitry is configured to:
generate the one or more transforms based on the one or more propagation characteristics; and
apply the one or more transforms to the one or more second samples.
50 . The apparatus of claim 26 , wherein the one or more channel characteristics comprise any of an estimated channel input response and an accumulation of estimated channel input response.
51 . A method comprising:
performing, at a receiver, radio frequency (RF) fingerprinting based at least in part on:
inputting, to a neural network, (i) one or more samples of a RF signal received from a transmitter and (ii) one or more channel characteristics of a channel between the receiver and the transmitter; and
obtaining a predicted value output from the neural network.
52 . An apparatus comprising:
circuitry, including any one or more of a processor and memory, the circuitry configured to perform radio frequency (RF) fingerprinting, wherein the RF fingerprinting comprises:
inputting, to a neural network, (i) one or more samples of a RF signal received from a transmitter and (ii) one or more channel characteristics of a channel between the apparatus and the transmitter; and
obtaining a predicted value output from the neural network.