Early Stage Internet Traffic Identification Using Data Gravitation Based Classification

Traditional machine learning traffic identification techniques usually use the features of a whole Internet flow, which makes such techniques few sense in engineering practices. Therefore, recent years, an increasing number of researchers turn to build effective machine learning models to identify traffics with the few packets at the early stage. In this paper, data gravitation based classification (DGC) model, a supervised learning approach inspired by Newton's universal gravitation law, is applied for early stage traffic identification. In the empirical study, two open data sets and a data set collected in our campus network are employed. Eight widely used supervised learning algorithms are compared with our approach in the identification experiments. Accuracy and Cohen's kappa coefficient are applied to evaluate the performances of compared methods. The experimental results suggest that DGC outperformed the other algorithms for most cases considering both of accuracy and kappa. Thus, DGC is effective for early stage traffic identification.