Comparison of Internet Traffic Identification on Machine Learning Methods

摘要

Traffic classification is the essential part in computer network. It can identify the traffic application so as to better manage the network, filter the insecure network flows and provide better network services. However, traditional traffic identification methods cannot work well when encounter opaque packets or more complex flows. Machine learning methods become the most efficient way to solve the problems existed in traditional ways, mainly including supervised learning and unsupervised learning. In this paper, two classic methods in supervised and unsupervised learning ways are applied to achieve the identification of abnormal traffic based on flow-level features. Besides, the comparison of training time, prediction time and the accuracy are given, which helps deep understand machine methods for traffic identification and design more efficient traffic identification solutions.