OpenCBD: A Network-Encrypted Unknown Traffic Identification Scheme Based on Open-Set Recognition

Figure 2

Block diagram of proposed method. (1) Closed-set individual training: randomly select a part of the known class data to train the individual model including the encoder and the classifier. 2) Closed-set ensemble training: using all known class data to train encoders and classifiers in individual models, where the weights of the encoders are locked. (3) Open-set testing: the ensemble model produces results to be identified. If the largest to-be-identified result is greater than or equal to the threshold, the test sample is classified into one of the classes; otherwise, it is classified as unknown.