DroneGuard: An Explainable and Efficient Machine Learning Framework for Intrusion Detection in Drone Networks

Aim:

          Design and deliver a lightweight, interpretable, and efficient intrusion detection framework that detects GPS-spoofing and Denial-of-Service (DoS) attacks in drone networks in (near) real time while producing human-readable explanations for each alarm.

Abstract:

       DroneGuard is an ML-based intrusion detection framework tailored for resource-constrained drone networks. It combines feature selection, data balancing (SMOTE), randomized search hyperparameter tuning, and a set of classical ML classifiers to detect GPS spoofing and DoS attacks. DroneGuard emphasizes explainability and aims for a tradeoff of high detection performance and low computational cost — with the Decision Tree (DT) model emerging as the best compromise for on-edge deployment.

Proposed System:

         The proposed system introduces an intelligent intrusion detection framework developed to safeguard drone networks against GPS spoofing and DoS attacks. It continuously monitors telemetry and communication data to identify deviations from normal patterns that may indicate intrusions. The framework includes preprocessing, feature selection, class balancing, and optimization steps to achieve high accuracy with minimal computation. For both GPS and DoS datasets, different machine learning algorithms like LightGBM (LGBM), Random Forest (RF), XGBoost (XGB), and Gradient Boosting (GBM)—were trained and compared to determine the most effective model for deployment. The best-performing model was then enhanced with explainability methods to highlight critical features contributing to each prediction. Overall, the proposed system offers a fast, interpretable, and efficient solution for securing real-time drone operations.

Advantage:

• Boosting models give higher accuracy and faster convergence than earlier ML models.
• They handle complex attack patterns efficiently with lower computation time.
• Models are lightweight, scalable, and suitable for real-time drone deployment.
• Integration with explainable AI improves transparency and feature understanding.