A Physics-Informed Residual and Particle Swarm Optimization Framework for Physics-Informed UAV GPS Spoofing...

Area of Science:

• Robotics and Control Systems
• Navigation and Positioning
• Machine Learning for Cybersecurity

Background:

• Unmanned aerial vehicles (UAVs) heavily depend on Global Navigation Satellite Systems (GNSS) for navigation.
• Global Positioning System (GPS) spoofing presents a critical vulnerability, compromising UAV navigation integrity.
• Existing detection methods often lack interpretability or robustness against diverse spoofing scenarios.

Purpose of the Study:

• To develop a robust and interpretable framework for detecting GPS spoofing attacks targeting UAVs.
• To enhance the reliability and security of UAV navigation systems against GNSS signal manipulation.
• To overcome the limitations of traditional feature engineering and black-box deep learning models in spoofing detection.

Main Methods:

• Integration of Physics-Informed Residual (PIR) modeling with Particle Swarm Optimization (PSO) and Extreme Gradient Boosting (XGBoost).
• Physically interpretable residual construction enforcing temporal and carrier level consistency across GNSS observables.
• PSO for global hyperparameter tuning to optimize classifier generalization and robustness.

Main Results:

• The PIR-PSO-XGBoost framework achieved a classification accuracy of 95.26% and an F1-score of 95.28% on a real-world GPS spoofing dataset.
• Demonstrated significant performance improvement over conventional machine learning baselines.
• Validated the effectiveness of physics-guided feature construction combined with swarm-optimized learning.

Conclusions:

• The proposed PIR-PSO-XGBoost framework offers a robust, efficient, and deployable solution for GPS spoofing detection in UAVs.
• Combining physics-informed feature engineering with optimized machine learning enhances detection capabilities.
• The approach improves interpretability and computational efficiency compared to existing methods.