Securing GPS Signals from Spoofing Attacks using Optimized Temporal Attention Recurrent Graph Convolutional Neural Network

Unmanned Aerial Vehicles (UAV) heavily rely on Global Positioning Systems (GPS) for navigation. Consequently, anti-spoofing technology is crucial for ensuring the safety of UAV operations. It is crucial to identify signal spoofing attacks in small unmanned aerial vehicles. This paper proposes a Secure GPS Signals from Spoofing Attacks using Optimized Temporal Attention Recurrent Graph Convolutional Neural Network (SGPS-SA-TARGCNN). Initially, the signals are collected from TEXBAT (Texas Spoofing Test Battery) dataset. Then, the signal is fed into pre-processing utilizing Regularized Bias-Aware Ensemble Kalman Filter (RBEKF) to normalize and clean the data. After pre-processing, the Temporal Attention Recurrent Graph Convolutional Neural Network (TARGCNN) is used to classify the types of spoofing attack as GPS spoofing attack and normal. Finally, Spider Wasp Optimization Algorithm (SWOA) is used to optimize the weight parameter of TARGCNN, which can precisely classify the type of GPS spoofing attack. The metrics like accuracy, precision, f1-Score, recall, specificity, RoC and computational time are considered. The SGPS-SA-TARGCNN achieves 26.36%, 20.69% and 30.29% high accuracy, 19.12%, 28.32% and 27.84% high precision, 12.04%, 13.45% and 22.80% higher call when compared with existing methods: Improved Machine Learning Ensemble Method for Securing Small UAV from GPS Spoofing Attacks (DNN-SUAV-GPSSA), Securing Autonomous Vehicles against GPS Spoofing Attacks: A Deep Learning Technique (SAV-GPSSA-CNN), and Deep POSE: Detecting GPS spoofing attack by Deep Recurrent Neural Network (DGPS-SA-DRNN) respectively.