Perturbation RF signal generator incorporating quantum computing with game theoretic optimization and related methods
A perturbation radio frequency (RF) signal generator is provided which generates a perturbed RF output signal to cause a signal classification change by an RF signal classifier. The perturbation RF signal generator may include a quantum computing circuit configured to perform quantum subset summing; and a processor. The processor may be configured to generate a game theory reward matrix for a plurality of different deep learning signal perturbation models, cooperate with the quantum computing circuit to perform quantum subset summing of the game theory reward matrix, select a deep learning signal perturbation model from the plurality thereof based upon the quantum subset summing of the game theory reward matrix, and generate the perturbed RF output signal based upon the selected deep learning signal perturbation model to cause the signal classification change in the RF signal classifier.
1 . A perturbation radio frequency (RF) signal generator to generate a perturbed RF output signal to cause a signal classification change by an RF signal classifier, the perturbation RF signal generator comprising:
a quantum computing circuit configured to perform quantum subset summing; and
a processor configured to
generate a game theory reward matrix for a plurality of different deep learning signal perturbation models,
cooperate with the quantum computing circuit to perform quantum subset summing of the game theory reward matrix,
select a deep learning signal perturbation model from the plurality thereof based upon the quantum subset summing of the game theory reward matrix, and
generate the perturbed RF output signal based upon the selected deep learning signal perturbation model to cause the signal classification change in the RF signal classifier.
2 . The perturbation RF signal generator of claim 1 wherein the game theory reward matrix includes rows corresponding to the different deep learning models, and columns corresponding to predictive deep learning metrics for the different deep learning models.
3 . The perturbation RF signal generator of claim 1 wherein the plurality of different deep learning signal perturbation models have different perturbation weights associated therewith.
4 . The perturbation RF signal generator of claim 1 wherein the processor is configured to generate perturbed signals for pairwise combinations of starting classes and target classes for each deep learning signal perturbation model, and generate the game theory reward matrix based thereon.
5 . The perturbation RF signal generator of claim 1 wherein the deep learning signal perturbation models comprise at least one of Fast Gradient Sign Method (FGSM), Carlini and Wagner, and Jacobian-based saliency map attack models.
6 . The perturbation RF signal generator of claim 1 further comprising a variational autoencoder (VAE) configured to encode RF input signal data to generate a latent vector, and decode the latent vector to generate new signal data; and wherein the processor is configured to generate the game theory reward matrix based upon the new signal data.
7 . The perturbation RF signal generator of claim 6 wherein the quantum computing circuit is configured to construct a Z-test for a test observation for each signal classification.
8 . The perturbation RF signal generator of claim 1 wherein the processor is configured to construct a 3D latent space for the deep learning signal perturbation models, generate clusters of eigenvalues from the deep learning signal perturbation models for different signal classes, and process the new signal data using the deep learning signal perturbation models for signal classification change detection based upon the clusters of eigenvalues.
9 . A perturbation radio frequency (RF) signal generator to generate a perturbed RF output signal to cause a signal classification change by an RF signal classifier, the perturbation RF signal generator comprising:
a quantum computing circuit configured to perform quantum subset summing; and
a processor configured to
generate a game theory reward matrix for a plurality of different deep learning signal perturbation models, the game theory reward matrix includes rows corresponding to the different deep learning signal perturbation models and columns corresponding to predictive deep learning metrics for the different deep learning models, and the plurality of different deep learning signal perturbation models having different perturbation weights associated therewith,
cooperate with the quantum computing circuit to perform quantum subset summing of the game theory reward matrix,
select a deep learning signal perturbation model from the plurality thereof based upon the quantum subset summing of the game theory reward matrix, and
generate the perturbed RF output signal based upon the selected deep learning signal perturbation model to cause the signal classification change in the RF signal classifier.
10 . The perturbation RF signal generator of claim 9 wherein the processor is configured to generate perturbed signals for pairwise combinations of starting classes and target classes for each deep learning signal perturbation model, and generate the game theory reward matrix based thereon.
11 . The perturbation RF signal generator of claim 9 wherein the deep learning signal perturbation models comprise at least one of Fast Gradient Sign Method (FGSM), Carlini and Wagner, and Jacobian-based saliency map attack models.
12 . The perturbation RF signal generator of claim 9 further comprising a variational autoencoder (VAE) configured to encode RF input signal data to generate a latent vector, and decode the latent vector to generate new signal data; and wherein the processor is configured to generate the game theory reward matrix based upon the new signal data.
13 . The perturbation RF signal generator of claim 9 wherein the processor is configured to construct a 3D latent space for the deep learning signal perturbation models, generate clusters of eigenvalues from the deep learning signal perturbation models for different signal classes, and process the new signal data using the deep learning signal perturbation models for signal classification change detection based upon the clusters of eigenvalues.
14 . The perturbation RF signal generator of claim 13 wherein the quantum computing circuit is configured to construct a Z-test for a test observation for each signal classification.
15 . A method for generating a perturbed radio frequency (RF) output signal to cause a signal classification change by an RF signal classifier, the method comprising:
using a processor to
generate a game theory reward matrix for a plurality of different deep learning signal perturbation models,
cooperate with a quantum computing circuit to perform quantum subset summing of the game theory reward matrix,
select a deep learning signal perturbation model from the plurality thereof based upon the quantum subset summing of the game theory reward matrix, and
generate the perturbed RF output signal based upon the selected deep learning signal perturbation model to cause the signal classification change in the RF signal classifier.
16 . The method of claim 15 wherein the game theory reward matrix includes rows corresponding to the different deep learning models, and columns corresponding to predictive deep learning metrics for the different deep learning models.
17 . The method of claim 15 wherein the plurality of different deep learning signal perturbation models have different perturbation weights associated therewith.
18 . The method of claim 15 further comprising using the processor to generate perturbed signals for pairwise combinations of starting classes and target classes for each deep learning signal perturbation model, and generate the game theory reward matrix based thereon.
19 . The method of claim 15 wherein the deep learning signal perturbation models comprise at least one of Fast Gradient Sign Method (FGSM), Carlini and Wagner, and Jacobian-based saliency map attack models.
20 . The method of claim 15 further comprising using a variational autoencoder (VAE) configured to encode RF input signal data to generate a latent vector, and decode the latent vector to generate new signal data; and using the processor to generate the game theory reward matrix based upon the new signal data.
21 . The method of claim 15 further comprising using the processor to construct a 3D latent space for the deep learning signal perturbation models, generate clusters of eigenvalues from the deep learning signal perturbation models for different signal classes, and process the new signal data using the deep learning signal perturbation models for signal classification change detection based upon the clusters of eigenvalues.
22 . The method of claim 21 further comprising using the quantum computing circuit to construct a Z-test for a test observation for each signal classification.