Encrypted Network Traffic Classification using Ensemble Learning Techniques

There is a continuous evolution of technological devices leading to a huge amount of traffic data on the internet. This presents Internet Service Providers with changes in the Quality of Service being provided and network security. The classification of these network traffic data promotes a better QoS, and management of the encrypted network. The major concern of the ISPs is protecting users' privacy, thereby generating network traffic data that are encrypted. In this thesis, we determine the best techniques as well as the relevant statistical features suitable for the classification of the non-VPN encrypted network traffic data. We utilize the opensource UNB and the Solana Networks encrypted network traffic datasets. We performed multiple experiments that led to developing an ensemble learning model with the stacking technique using the deep learning and machine learning classification algorithms with the best performances for the classification of the non-VPN encrypted network traffic data.