GPS Spoofing Detection Method for Small UAVs Using 1D Convolution Neural Network.

Area of Science:

• Robotics and Autonomous Systems
• Cybersecurity
• Artificial Intelligence

Background:

• Small unmanned aerial vehicles (UAVs) heavily rely on global positioning system (GPS) for navigation.
• UAVs are susceptible to GPS spoofing attacks, which can compromise their operational safety.
• Existing anti-spoofing methods often require additional hardware, posing challenges for capacity-limited small UAVs.

Purpose of the Study:

• To propose a novel, lightweight, and power-efficient deep learning-based anti-spoofing method for UAVs.
• To enable real-time GPS spoofing detection on mobile drone platforms without additional hardware.
• To enhance the performance of the anti-spoofing technique through increased training data and network architecture adjustments.

Main Methods:

• Development of a deep learning model utilizing a 1D convolutional neural network (CNN).
• Implementation and evaluation of the algorithm on an embedded drone platform.
• Comparison of the proposed method against a support vector machine (SVM) using precision, recall, and F-1 score metrics.

Main Results:

• The proposed 1D CNN method demonstrated superior performance compared to SVM in detecting GPS spoofing.
• The algorithm proved to be lightweight and power-efficient, suitable for real-time application on mobile platforms.
• Flight tests confirmed the successful detection of GPS spoofing attacks by the developed algorithm.

Conclusions:

• The deep learning-based anti-spoofing method is effective and suitable for small UAVs.
• The approach offers a viable solution for enhancing UAV navigation security against GPS spoofing.
• Further improvements in performance are achievable by optimizing training data and network architecture.