Algorithm 1. AE as one-class classifier.

Input: X-Training and Testing Set

Output: Classification Results

Initialize: all AE network parameters

Procedure:

Phase-I: Training

for each training epoch do

for each mini batch do

1. H ← Feature representation using Eq. (1)

2. Z ← Reconstructed Input using Eq. (2)

3. Compute the gradient to minimize Cost function in Eq. (3) using Adam

4. Update AE network parameters

end for

end for

α ← average Reconstruction Error on Training data set

Phase-II: Testing

for each sample in Testing dataset do

1. Error ← Compute reconstruction Loss

2. if Error > α then Sample is attack else Sample is normal traffic

end for