Encrypted Network Traffic Classification Using a Geometric Learning Model

The increasing volume of encrypted traffic from emerging applications has made conventional traffic classification approaches ineffective and called for novel methods for identifying network flows. Recent machine learning and deep learning based approaches have been proposed yet many of them are severely limited by their feature selection and inherent neural network architecture. As network data by nature are of non-Euclidean distance space and carry abundant chronological relationship, we are inspired to utilize geometric deep learning that simultaneously takes into account packet raw bytes, metadata and packet relations for classifying encrypted network traffic. We validate our proposed graph neural network (GNN) models against the reference methods including convolutional neural networks (CNN) and recurrent neural networks (RNN) quantitatively and demonstrate that the proposed graph neural networks outperform the state of art.