A Multi-Level Network Traffic Classification in Combating Cyberattacks Using Stack Deep Learning Models
Believe Ayodele, Victor Buttigieg · 2024
Recent advances in computing power and the growing number of threats to IT infrastructure have increased the use of Deep Learning (DL) models in combating attacks due to their ability to act as an inference engine in detecting and classifying cyberattacks. However, the DL models proposed so far fall short by focusing predominantly on a single type of network traffic analysis, which is usually dictated by the dataset used. This research presents a stack model of Recurrent Neural Networks (RNN), Long-Short-Term memory (LSTM), and Gated Recurrent Networks (GRU) to combat cyberattacks by utilising multilevel traffic analysis at the packet, flow, and session levels. The proposed RNNLSTM-GRU, RNNGRU-LSTM, and LSTMGRU-RNN models are exploited on the same dataset to compare performance. The results show that the models performed well as binary and multiclassifiers, with an accuracy of up to 99.99%. With flow and session-level training, the performance was somewhat lower, achieving accuracies of up to 97.63%.