DeepSpoofNet: a framework for securing UAVs against GPS spoofing attacks.

Area of Science:

• Cybersecurity
• Aerospace Engineering
• Artificial Intelligence

Background:

• Uncrewed Aerial Vehicles (UAVs) are increasingly vital in various sectors, but susceptible to Global Positioning System (GPS) spoofing attacks.
• GPS spoofing can compromise UAV control, leading to capture or destruction, necessitating robust detection mechanisms.
• Existing machine learning methods face challenges like imbalanced datasets and limited accuracy in resource-constrained environments.

Purpose of the Study:

• To investigate optimal feature selection (FS) and deep learning techniques for detecting GPS spoofing attacks on UAVs.
• To address limitations in current research, including imbalanced datasets and sub-optimal feature selection.
• To develop a framework that improves the accuracy and reliability of GPS spoofing detection in UAVs.

Main Methods:

• Implemented rigorous oversampling techniques to handle imbalanced datasets.
• Combined advanced feature selection methods with a hybrid neural network (NN) architecture.
• Utilized a Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) integrated model.

Main Results:

• The proposed Analysis of Variance (ANOVA) + CNN-BiLSTM hybrid model achieved superior performance.
• Exceptional results include 98.84% precision, 99.25% accuracy, 99.26% F1 score, and 99.69% recall.
• Demonstrated significant improvements in prediction accuracy, true/false positive rates, F1 score, and recall.

Conclusions:

• The developed hybrid model effectively detects GPS spoofing attacks on UAVs, overcoming existing research challenges.
• The framework offers a robust solution for enhancing UAV security against sophisticated cyber threats.
• The findings pave the way for more secure and reliable autonomous navigation systems.