Table 5.

Hyperparameter details for ImmuneNet for CIC IDS 2017, CIC IDS 2018, and CIC Bell DNS 2021 datasets.

Hyperparameters	Values as per dataset
	CIC IDS 2017	CIC IDS 2018	CIC bell DNS 2021
Input feature dimension	38	38	32
Batch size	64	64	64
Adam optimizer	β1 = 0.9, β2 = 0.99, ε = 1e-7	β1 = 0.9, β2 = 0.99, ε = 1e-7	β1 = 0.9, β2 = 0.99, ε = 1e-7
Learning rate	1e-4	1e-4	3e-4
Additional optimization strategies	Cosine annealing learning rate scheduling, early stopping	Cosine annealing learning rate scheduling, early stopping	Cosine annealing learning rate scheduling, early stopping
# Epochs	15	15	15
# Units in input FC layer	512	512	512
Input FC layer activation	Mish	Mish	Mish
# Units in hidden FC layer 1	784	784	784
FC layer 1 activation	Mish	Mish	Mish
# Units in hidden FC layer 2	512	512	512
FC layer 2 activation	Mish	Mish	Mish
# Units in output FC layer (binary classification)	1	1	1
Output FC layer activation	Sigmoid	Sigmoid	Sigmoid