| The continuously increasing number of activities processed via the internet, often leaves the user vulnerable to cyber-attacks. The goal of the scientific community is to deploy innovative approaches and methodologies, capable to offer protection from potential cyber threats. This research effort aims to contribute to networks’ security by introducing the Hybrid Ensemble Deep Learning (HEDL) Intrusion Detection System (IDS) that successfully detects nine serious cyber-attacks. Its architecture comprises of three Deep Neural Networks (DNN), three Convolutional Neural Networks (CNN) and 3 Recurrent Neural Networks (RNN) using Long-Short Term Memory (LSTM) layers, running in parallel. The HEDL-IDS was successfully tested against the UNSW-NB15 dataset, achieving an overall accuracy of 98.35% and 96.25% in the training and testing phases respectively. The performance of the proposed model was evaluated by calculating Accuracy, Sensitivity, Specificity, Precision and F-1 Score. The values of all above indices were higher than 0.92, indicating the accurate performance of the developed model. The HEDL-IDS was compared with 20 robust Machine Learning Classification algorithms, sealing its reliability. |
*** 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.