Scholarlink

<p>The safety of Global Positioning System (GPS) is pivotal for various navigation and location-based applications. Nevertheless, GPS spoofing attacks significantly affect the accuracy and reliability of GPS signals. In this study, an innovative hybrid model for detecting spoofed GPS, entitled EKF- Enhanced Attention-based CNN-GRU (ACGN) is introduced. To promote noise reduction and signal smoothing, the model requires implementation of an extended Kalman filter (EKF) for preprocessing GPS signal data before further analysis. Spatial dependencies in the filtered data is learned by convolutional neural networks (CNN), which are also added to the Gated Recurrent Units (GRU) to enable more effective modeling of temporal characteristics of the signal. Feature representation is further enhanced with the Convolutional Block Attention Module (CBAM) by instantiating channel and spatial attention mechanisms, making it possible for the model to differentiate between true and spoofed. The results show that detection accuracy is 99.4% with very low Detection Error Rate (DER) of approximately 0.03. The work hence contributes and provide real-time solution to this problem in completely enhancing GPS-based system security and integrity.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel EKF-integrated attention-enhanced CNN-GRU framework for precise GPS spoofing detection

  • 期刊论文

  • 发表日期：

    2025年4月16日

查看全文
  • J. Cynthia,
  • S. Rathi

摘要

The safety of Global Positioning System (GPS) is pivotal for various navigation and location-based applications. Nevertheless, GPS spoofing attacks significantly affect the accuracy and reliability of GPS signals. In this study, an innovative hybrid model for detecting spoofed GPS, entitled EKF- Enhanced Attention-based CNN-GRU (ACGN) is introduced. To promote noise reduction and signal smoothing, the model requires implementation of an extended Kalman filter (EKF) for preprocessing GPS signal data before further analysis. Spatial dependencies in the filtered data is learned by convolutional neural networks (CNN), which are also added to the Gated Recurrent Units (GRU) to enable more effective modeling of temporal characteristics of the signal. Feature representation is further enhanced with the Convolutional Block Attention Module (CBAM) by instantiating channel and spatial attention mechanisms, making it possible for the model to differentiate between true and spoofed. The results show that detection accuracy is 99.4% with very low Detection Error Rate (DER) of approximately 0.03. The work hence contributes and provide real-time solution to this problem in completely enhancing GPS-based system security and integrity.