Table 3 The network structure of the generator and the discriminator in LSGAN
From: LSGAN-AT: enhancing malware detector robustness against adversarial examples
The generator (G) network structure | The discriminator (D) network structure | ||
|---|---|---|---|
Layer type | Output shape | Layer type | Output shape |
Input Layer (Malware) | (None, 128) | Input Layer | (None, 128) |
Input Layer (Noise) | (None,20) | LeakyReLU(Dense) | (None,256) |
Concatenate (Malware + Noise) | (None,148) | Dropout(0.05) | (None,256) |
LeakyReLU (Dense) | (None,256) | LeakyReLU(Dense) | (None,256) |
Batch Normalization | (None,256) | Dropout(0.05) | (None,256) |
LeakyReLU(Dense) | (None,256) | LeakyReLU(Dense) | (None,256) |
Batch Normalization | (None,256) | Dropout(0.05) | (None,256) |
LeakyReLU (Dense) | (None,256) | Sigmoid(Dense) | (None,1) |
Batch Normalization | (None, 256) | Â | Â |
Sigmoid (Dense) | (None,128) | Â | Â |
Maximum | (None,128) | Â | Â |