RFGRU: A novel approach for mobile application traffic identification

Abstract

Billions of users access the Internet through their mobile devices to get services. Mobile traffic classification has become a hot topic in recent years due to its large volume of traffic data. Many of the studies that have been done show that the key point of mobile traffic identification is to extract signatures. However, the process of signature extraction is usually too complex to perform. In this paper, we propose a novel method RFGRU which is based on the Random Forest and gated recurrent unit, to address the mobile traffic classification problem. Several experiments are performed to verify the effectiveness of RFGRU. The results show that RFGRU delivers a good recognition rate and can accurately identify the traffic of the mobile applications.