On the fly classification of traffic in Anonymous Communication Networks using a Machine Learning approach

Abstract

Anonymous Communication Networks (ACNs) provide privacy and anonymity to the users of the Internet. Traffic classification in ACNs is an emerging area of research due to its benefits in network management tasks like network security, Quality of Service provisioning, and in Research and Development of ACNs. Out of the well-known traffic classification approaches available, Machine Learning (ML) based approach has proven to be advantageous over the port-based and payload based approach. Using a publicly released Anon17 dataset, this work presents an ML-based traffic classification technique in ACNs. The proposed technique performs on the fly classification, which involves the classification of traffic as early as possible using the first few packets of traffic flow. The proposed on the fly classification technique outperforms the state of the art technique in ACNs with increased classification accuracy, F measure and requires less number of packets in traffic flow to achieve highest possible performance metrics.