Multi-Layer AI Sensor System for Real-Time GPS Spoofing Detection and Encrypted UAS Control.

Area of Science:

• Aerospace Engineering
• Artificial Intelligence
• Cybersecurity

Background:

• Unmanned Aerial Systems (UAS) are vital in civilian and defense sectors.
• Reliance on unencrypted GPS signals makes UAS vulnerable to spoofing attacks, threatening safety and operations.

Purpose of the Study:

• To introduce an AI-driven, multi-layer sensor framework for real-time GPS spoofing detection and secure command-and-control (C2) in UAS.
• To enhance telemetry reliability and enable secure communication in resource-constrained UAS platforms.

Main Methods:

• A refined preprocessing pipeline with GPS Drift Index (GDI), statistical normalization, oversampling, Kalman filtering, and quaternion filtering.
• Differentiable Architecture Search (DARTS) for generating lightweight neural networks for onboard spoofing detection.
• PRESENT-128 encryption and CMAC authentication for secure C2 communication.

Main Results:

• The framework achieved outstanding detection accuracy (99.99%), F1-score (0.999), and AUC (0.9999).
• Secure C2 communication with low latency (1.79 ms) and energy cost (0.51 mJ).
• Demonstrated suitability for real-world, resource-constrained UAS environments.

Conclusions:

• The AI-driven framework provides a robust, scalable, and secure solution for countering GPS spoofing in autonomous aerial vehicles.
• This research advances AI-enabled sensor systems for enhanced UAS navigation and security.