I. Introduction
The identification of network applications plays an important role in network management [1], [2], which has been one of the most popular research fields in network security [3], [4]. The continuously developing techniques in network application identification are widely employed to advanced network protection, including detecting malicious traffic with Intrusion Detection Systems (IDS) [5], [6], [7] and monitoring network behavior with intelligent gateway [8], [9], etc. Encrypted network traffic has gradually become the mainstream since the prosperity of encryption technology. Identification methods based on rule matching reach their limitation when tackling encrypted traffic. The newly emerged side-channel methods with artificial intelligence offer a better way to classify encrypted traffic [4], [10], [11], [12], [13]. Theoretically, there are four main types of state-of-the-art encrypted traffic methods.