A Sequential Hybrid Intrusion Detection Model for Detecting DoS Attacks in UAV Networks

The significant research contribution is the design of a sequential hybrid IDS (MLP with Random Forest) for detecting DoS attacks in UAV networks. The model combines the MLP’s ability to learn complex nonlinear traffic patterns with Random Forest’s ensemble-based decision refinement. Using 5-fold out-of-fold stacking and SMOTE-based class balancing, the proposed approach achieved 0.92 accuracy, 0.95 recall, 0.84 F1-score, and approximately 0.935 ROC-AUC, outperforming the evaluated standalone baseline models.