FusionTC: Encrypted App Traffic Classification Using Decision-Level Multimodal Fusion Learning of Flow Sequence

Algorithm 1

Stacking.

Input: Training data

Output: An ensemble classifier H

1: Step1: Train base-classifiers at first-level

2: form ←1 to Mdo

3: Train H_m using D

4: end for

5: Step2: Construct new dataset U_train using D

6: U_train = null

7: fori ←1 to ndo:

8: form ←1 to Mdo:

9:  z_it = H_m(x_i)

10: end for

11: U_train = U_train∪{(z_1i,z_2i,⋯,z_Mi ),y_i}

12: end for

13: Step3: Train the meta-classifier at second-level

14: Train the meta-classifier L using U_train

15: returnH = { H₁, H₂,…H_i, …H_M , L}