FORTIS: A Hybrid Ensemble Model for Industrial Cybersecurity and Intrusion Detection

Abstract

Cyber-attacks are becoming more advanced, making it harder to detect malicious network activity. Because of this, intrusion detection systems (IDS) need to provide high detection accuracy while keeping false alarms low. This paper compares different deep learning models for network intrusion detection and introduces FORTIS, a hybrid ensemble model for industrial cybersecurity and intrusion detection. The proposed model combines the prediction scores of a deep neural network (DNN), a one-dimensional convolutional neural network (1D-CNN), XGBoost, and Random Forest using weighted fusion based on validation results. The model is evaluated on the NSL-KDD dataset, which contains 125,973 labeled network connection records with 41 traffic features. Using a stratified $\mathbf{8 0} / \mathbf{2 0}$ train test split, the 1D-CNN and DNN achieve test accuracy of 97.87% and 99.51%, respectively. The proposed FORTIS model achieves the best performance with 99.92% accuracy, a 99.91% F1-score, and a 100.00% ROC-AUC. The results show that combining deep learning and tree-based models improves detection performance compared with using a single model. This makes FORTIS a reliable and effective approach for industrial cybersecurity and network intrusion detection.