The growing volume of online activities exposes users to potential cyber-attacks. Consequently, the scientific community aims to develop pioneering approaches capable to mitigate the risk. To address this challenge, the au-thors introduce the second version of the Hybrid Ensemble Deep Learning (HEDL) Intrusion Detection System (IDS), that successfully detects nine se-rious cyber-attacks. The architecture of the introduced Ensemble comprises of four Deep Neural Networks (DNN), four Convolutional Neural Networks (CNN) and four Recurrent Neural Networks (RNN) using Long-Short Term Memory (LSTM) layers, running in parallel. The final classification of each Ensemble employs a Custom Vote process, following the Weighted Vote and the Majority Vote principles. The HEDL-IDS2 was successfully employed on the UNSW-NB15 dataset, achieving extremely high-performance indices (Ac-curacy, Sensitivity, Specificity, Precision and F-1 Score) in all Training, Val-idation and Testing phases. This multiclass classification effort followed the One-Versus all Strategy. |
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.