A Comparative Analysis of Ensemble-Based Machine Learning Approaches with Explainable AI for Multi-Class Intrusion Detection in Drone Networks

Drone networks face multiple sophisticated cyber threats that require specialized detection systems. The research identifies nine distinct intrusion types targeting drones, including spoofing, injection, replay, and man-in-the-middle attacks, highlighting that generic cybersecurity approaches may be insufficient for this emerging technology.

Machine learning can achieve near-perfect detection accuracy for drone intrusions. Random Forest models reached a macro F1-score of 0.9998 and ROC AUC of 1.0000, suggesting that ensemble-based approaches are highly effective at identifying and classifying different types of network attacks on drone systems.

Explainability is being integrated into intrusion detection systems. The research emphasizes that models aren't just accurate—they also incorporate explainable AI methods, meaning security teams can understand why the system flagged a particular network activity as suspicious, rather than treating it as a "black box" decision.

Rigorous statistical validation strengthens confidence in AI security systems. The study applied multiple statistical tests (Friedman's test, Wilcoxon signed-rank test, bootstrapped confidence intervals) to validate model performance, demonstrating that responsible AI development for cybersecurity requires more than just accuracy metrics—it demands robust evidence that results are reliable and reproducible.

Ensemble methods outperform individual approaches. By comparing multiple machine learning models (Random Forest, XGBoost, CatBoost, AdaBoost, Extra Trees), the research shows that combining different algorithms provides better protection than relying on a single detection method.