Dynamic Selection Techniques for Detecting GPS Spoofing Attacks on UAVs.

Area of Science:

• Cybersecurity
• Aerospace Engineering
• Machine Learning

Background:

• Unmanned aerial vehicles (UAVs) face significant cyber threats, particularly Global Positioning System (GPS) spoofing.
• Existing detection methods are insufficient, as evidenced by frequent GPS spoofing incidents.
• There is a critical need for robust security solutions to protect UAVs.

Purpose of the Study:

• To propose novel dynamic selection techniques for identifying effective classifiers in GPS spoofing detection.
• To develop a one-stage ensemble feature selection method for data preprocessing.
• To enhance the security and reliability of UAVs against GPS spoofing attacks.

Main Methods:

• Developed two dynamic selection techniques: Metric Optimized Dynamic selector and Weighted Metric Optimized Dynamic selector.
• Implemented a one-stage ensemble feature selection to refine the dataset.
• Evaluated ten machine learning models using accuracy, probability of detection, false alarm, misdetection, and processing time.

Main Results:

• The proposed dynamic techniques achieved superior performance compared to existing ensemble models.
• Achieved high accuracy (99.6%) and probability of detection (98.9%).
• Demonstrated low probability of false alarm (1.56%) and misdetection (1.09%) with a processing time of 1.24 s.

Conclusions:

• The novel dynamic selection techniques effectively identify optimal classifiers for GPS spoofing detection in UAVs.
• These methods offer a significant advancement in securing UAVs against sophisticated cyber-attacks.
• The proposed approach provides a reliable and efficient solution for enhancing UAV cybersecurity.