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

The encryption of network traffic promotes the development encrypted classification and identification research. However, many existing studies are only effective for closed-set experimental data, that is to say, known classes, while there often lots unknown classes in real environment open sets, have difficulty identifying can misclassify them as classes. How identify classify an open-collection one focuses analysis Considering these problems, this paper proposes a novel solution, which applies open-set recognition method identification, constructs model based on deep learning ensemble learning. convolutional neural transformer encoder then uses three-stage training testing process, combined with loss function, generalize space form OpenCBD. Experiments public datasets show proposed significantly better than other methods. It not distinguish from but also specific traffic.