Abstract: A robust network traffic classification and new type discovery algorithm is proposed by this paper, which is based on sparse autoencoder to extract feature features and classify based on threshold-based active learning classification algorithm. In addition, to achieve the purpose of identifying new application types, the excellent performance of the proposed algorithm through comparative experiments is verified. Among them, the accuracy of the classification algorithm can reach 91.08%; the recognition of new application types can reach 98.8%.

Key words: sparse self-encoder, feature extraction, active learning, robustness, new type discovery