A Method for Detecting Abnormal Traffic in Full-Stream Network Based on Machine Learning Technology

EasyChair Preprint 3405

Abstract

For the network, each server computer, and even the terminal system, abnormal network traffic will cause a lot of CPU time slices and memory space occupation, and cannot respond to demand services normally. In order to solve these problems, it is necessary to build an analysis system of network traffic anomaly, which has good functions of early warning, alarm and traffic processing. This paper proposes a full-flow network abnormal traffic detection method based on machine learning technology, using machine learning technology as a classifier and interpreter to detect abnormal traffic data in the network and output a conclusion report. By importing the network traffic data intercepted from the network into the database, extracting relevant data from the database, constructing a data frame and data point collection, and designing a unique data conversion mechanism for the data, and finally detecting the data points in the data frame and classification and other operations, to obtain the analysis and explanation of normal data, abnormal data and abnormal behavior after classification, and output data analysis static report.

Keyphrases: Full Flow Detection, Network Abnormal Traffic, machine learning

@booklet{EasyChair:3405,
  author    = {Yutong Han and Huaibin Wang and Jiongming Zhu},
  title     = {A Method for Detecting Abnormal Traffic in Full-Stream Network Based on Machine Learning Technology},
  howpublished = {EasyChair Preprint 3405},
  year      = {EasyChair, 2020}}