Exploring Multi-Channel GPS Receivers for Detecting Spoofing Attacks on UAVs Using Machine Learning.

Area of Science:

• Navigation Systems
• Cybersecurity
• Aerospace Engineering

Background:

• Global Positioning System (GPS) is crucial for transportation navigation but vulnerable to spoofing attacks.
• GPS spoofing can lead to loss of position monitoring, increasing risks of crashes and hijacking.
• Existing research often focuses on single attack types, necessitating a broader approach.

Purpose of the Study:

• To develop and evaluate machine learning models for detecting and classifying diverse GPS spoofing attacks on unmanned aerial vehicles (UAVs).
• To propose an interpretable data processing method for UAV GPS signal analysis.

Main Methods:

• Reviewed existing GPS spoofing detection and mitigation techniques.
• Processed a dataset of authentic and spoofed UAV GPS signals, extracting mission sequences and structuring data.
• Applied tree-based machine learning algorithms: Decision Tree, Random Forest, and XGBoost for classification.

Main Results:

• Random Forest demonstrated superior capability in detecting and classifying GPS spoofing attacks compared to other models.
• The study successfully identified and distinguished most types of GPS spoofing attacks.
• The proposed interpretable approach enhanced the analysis of complex GPS signal data.

Conclusions:

• Machine learning, particularly Random Forest, offers a robust solution for identifying GPS spoofing in UAVs.
• A comprehensive approach analyzing multiple attack types is essential for effective mitigation.
• The developed methods contribute to enhancing the security and reliability of UAV navigation systems.