Emilie: Enhance the power of traffic identification

摘要: Network traffic identification has become more and important in recent years. However, as the Internet backbone bandwidth continuously grows, traditional flow-based methods gradually impractical. In order to improve performance of identification, this paper proposes an ingenious practical flow dispatching mechanism named Emilie, which intelligently predicts elephant flows using only first three packets each flow. By discriminating mouse against flows, with various complexity are utilized identify application-level protocol type separately. Emilie utilizes Machine Learning techniques achieve high accuracy well keep fast speed predicting flows. Experimental results on real network traces illustrate that around 88% precision, 85% recall over gained average, is much better than existing solutions. To best our knowledge, efficient work supports inline prediction. Flow based empowers systems both speed.