Lightweight Tor flow classification method and system based on quaternary feature fusion graph

• Main Authors: SONG XIANGMEI, HE LIUKUN, ZHAO HUI, ZHOU QIANG, WANG LIANGMIN, LI CHANG, SHEN TUHAO
• Format: Patent
• Language: Chinese; English
• Published: 15.03.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a lightweight Tor flow classification method and a lightweight Tor flow classification system based on a quaternary feature fusion graph, which are used for classifying Tor flow generated by a network user by combining an anonymous network Tor multiplexing characteristic and a requirement on a model computing power in actual deployment. Firstly, a target Tor traffic set is captured from a test environment, then traffic in a fixed window is converted into a quaternary feature fusion graph, then fusion graph features are extracted by using a stacked feature extraction layer sharing partial parameters, and the features are refined and dimensionally reduced by using two different pooling strategies respectively, so that the target Tor traffic set is obtained. And finally, using the trained known category judgment model and unknown category judgment model to distribute category labels for the feature vectors. According to the method, efficient classification of the Tor traffic and lightweig