DarkNet Traffic Classification Pipeline with Feature Selection and Conditional GAN-based Class Balancing
Sashank Sridhar, Sowmya Sanagavarapu · 2021
DarkNet is an encrypted collection of internet sites that host criminal activities and hidden services. These privatized networks keep the internet activity anonymous and almost untraceable. DarkNet traffic classification aids in enhancing the security of any network through detection of threats or risks to any network of systems. In this paper, the standard CIC-Darknet2020 dataset used contained instances of benign and DarkNet traffic to a network. Feature importance analysis is performed using Chi-Squared statistical score on the dataset to aid in feature selection. The imbalance of the classes is then handled by performing oversampling using Conditional Generative Adversarial Networks. The multi-class classification of the traffic encryption type is performed using Random Forest classifier. This pipeline performs with a F1-Score of 97.87 for traffic encryption classification.