Identifying IoT Devices Based on Spatial and Temporal Features from Network Traffic

Algorithm 2

Training process of the Conv-BiLSTM model.

Input:
	composed of network flows, the dimension of each network flow is 2500.
	{, , , } represent some of the parameters during model training.
Output:
	The categories of
(1)	for each epoch in (1, ) do
(2)	for each data of the do
(3)	for each in batch do
(4)	Reshape to 50 50 form
(5)	Compute convolution with 6 filters
(6)	Compute the result through
(7)	Max Pooling
(8)	Compute convolution with 16 filters
(9)	Compute the result through
(10)	Max Pooling
(11)	Flatten the data
(12)	Run through a densely connected layer
(13)	Dropout
(14)	Reshape output data as 10 ∗ 160
(15)	Run through the 2-layered BiLSTM with dropout
(16)	Run through a densely connected layer
(17)	Output the result referring
(18)	Update the parameters of weight and bias
(19)	end for
(20)	end for
(21)	end for