November 2019
Intermediate to advanced
346 pages
9h 36m
English
We start by importing all the MalGAN code that we will be using (Step 1). We must specify a few arguments (Step 2), to be explained now. The savePath parameter is the location into which the adversarial examples will be saved. The modelPath variable is the path to the pre-computed weights of MalConv. The logPath parameter is where data pertaining to the application of the Fast Gradient Signed Method (FGSM) to the sample is recorded. For example, a log file may appear as follows:
|
filename |
original score |
file length |
pad length |
loss |
predict score |
|
0778...b916 |
0.001140 |
235 |
23 |
1 |
0.912 |
Observe that the original score is close to 0, indicating that the original sample is considered malicious by MalConv. ...