Real Network Traffic Collection and Deep Learning for Mobile App Identification

Table 1

Hyperparameters of the proposed DL models.

Hyperparameters	SDAE	1D CNN	LSTM
Input units	300	1014	20
Hidden layer units	300,200,100	—	100
Number of layers	7	8	5
Dropout rate	0.05	0.5	0.5
Activation	Tanh	ReLU	ReLU
Optimizer	Adam	Adam	Adam
Batch size	30	30	30
Training epochs	20	20	100
Number of filters	—	256	—
Filter size	—	7	—
Pooling size	—	3	—