Incremental High Throughput Network Traffic Classifier

H.R. Loo, Alireza Monemi, Trias Andromeda, M. N. Marsono


Today’s network traffic are dynamic and fast. Con-ventional network traffic classification based on flow feature and data mining are not able to process traffic efficiently. Hardware based network traffic classifier is needed to be adaptable to dynamic network state and to provide accurate and updated classification at high speed. In this paper, a hardware architecture of online incremental semi-supervised algorithm is proposed. The hardware architecture is designed such that it is suitable to be incorporated in NetFPGA reference switch design. The experimental results on real datasets show that with only 10% of labeled data, the proposed architecture can perform online classification of network traffic at 1Gbps bitrate with 91% average accuracy without loosing any flows.

Full Text: PDF


  • There are currently no refbacks.