Fig 1.
AES-128 structure.
Fig 2.
Profiling side-channel attack.
Fig 3.
Inception block parameter change process.
Fig 4.
Process of feature fusion.
Fig 5.
A single Inception module structures the feature extraction process.
Fig 6.
InceptionNet model.
Table 1.
Hyperparameter selection.
Fig 7.
The basic architecture of an autoencoder.
Fig 8.
Long short-term memory network structure diagram.
Fig 9.
LU-Net model.
Fig 10.
Comparison of attack results on the ASCAD dataset.
A: Training Accuracy B: Attack Performance.
Table 2.
Comparison of training parameters of ASCAD dataset.
Fig 11.
Comparison of attack results on the DPA contest v4 dataset.
A: Training Accuracy B: Attack Performance.
Table 3.
Comparison of training parameters of DPA contest v4 dataset.
Fig 12.
Comparison of attack results on the AES_RD dataset.
A: Training Accuracy B: Attack Performance.
Table 4.
Comparison of training parameters of AES_RD dataset.
Fig 13.
Using LU-Net model to remove gaussian noise.
Fig 14.
A: Denoised with DAE B: Denoised with LU-Net.
Fig 15.
Using LU-Net model to remove permutation noise.
Fig 16.
A: Denoised with DAE B: Denoised with LU-Net.
Table 5.
Comparison of denoising performance between LU-Net and DAE models.