Robust ConvNet-Kalman Filter Integration for Mitigating GPS Jamming and Spoofing Attacks Basing on Inertial Navigation System Data

GPS (Global Positioning System) is the most accurate system for various applications, especially in transportation. However, GPS is critically vulnerable due to its reliance on radio signals, which can be exploited by hackers through intentional attacks like spoofing and jamming, leading to potentially dangerous disruptions for both humans and services. Moreover, GPS systems can also experience accidental disruptions in urban environments, where signals from multiple satellites may be blocked by buildings, severely affecting the receiver's accuracy. This paper presents a robust method designed to mitigate GPS outages caused by both jamming and spoofing by integrating inertial data. The proposed method leverages two key components: convolutional neural networks (ConvNet) and the Kalman filter (KF). A carefully optimized deep layer in the ConvNet is employed to correct errors in the inertial navigation system (INS). The findings indicate a considerable enhancement in accuracy, with the proposed method reducing the RMSE by 77.68% compared to standalone GPS and by 98.34% compared to standalone INS. This significant improvement underscores the proposed approach's performance in maintaining reliable navigation in environments where GPS signals are compromised