A Novel Deep Encrypted Network Traffic Discriminator in Software Defined Network (SDN)
Negin Mohammadi, Alireza Shirmarz · Research Square · 2022
Abstract Nowadays, Internet users are rising and need to be supplied with an adoptable quality of service (QoS). Network traffic classification is one of the essential functions that can lead the internet service provider (ISP) to provide required network resources rationally. In facing new flows, the network traffic classification accuracy improvement can play a critical role in network performance, QoS, and security improvement. In this paper, we propose a novel classification model, including (1) a deep autoencoder and (2) a classifier to improve the network traffic classification accuracy in facing new network flows. The deep autoencoder is designed and evaluated in this article with the mean square error (MSE) metric. The proposed deep autoencoder has advanced the model to extract the effective features from the training set more accurately than other methods like the manual method or shallow neural network model. Three distinct classifiers are considered to be added to the deep autoencoder and make it more accurate. The transfer learning is used to add the distinct classifiers, namely logistic regression, random forest, decision tree, and Support Vector Machine (SVM), as a layer to the proposed model. The proposed deep classification model is evaluated with accuracy and f-score measures. The simulation results show that the proposed model has more accuracy and f-score than Convolutional Neural Network (CNN). The UNB ISCX VPN-nonVPN dataset is used for training and testing the model. Software Defined Network (SDN) architecture is used for the proposed model to be deployed because this architecture has made the network more programmable and flexible than the traditional closed networks.