DarkNet Traffic Recognition Using Meta-Learning

Darknet is a private network where communication between peers occurs only over communications protocols and non-standard ports. Darknets differ from distributed peer-to-peer networks in that they are anonymously shared, meaning that the IP address is not generally shared. Transaction is done through specialized tools such as TOR, allowing anonymity of the activities taking place. This paper proposes DarkNET traffic detection using ensemble learning and proposes an ensemble learning architecture based on stacking. The CIC-Darknet2020 dataset is the primary tool on which the models are built and tested. The comparative results showed the superiority of group learning and that the suggested model, which reached 98.8% accuracy, was the most accurate, which confirms the importance of using it in modeling network traffic data and distinguishing the DarkNET.