Scholarlink

<p>Global Navigation Satellite Systems (GNSS) play a crucial role in contemporary life, enabling various applications such as autonomous vehicles, agricultural industries, marine navigation, and military operations. However, the performance of GNSS receivers is susceptible to both intentional and unintentional interferences, with satellite spoofing attacks posing a significant threat. This work presents a GNSS detector receiver implemented using an STM32H7 series ARM microcontroller and a U-blox ZED-F9T receiver. Leveraging decision-making methods such as Fuzzy Logic (FL), Multi-Layer Perceptron Neural Network (MLP NN), and Random Forest (RF), as proposed in recent literature, the detector receiver aims to effectively detect spoofing attacks. In addition, the Constrained Complexity Minimization (CCM) algorithm was applied to compromise the complexity and accuracy of the RF Machine Learning (ML) model. According to the results, with the fine-tuned configuration, the optimized RF model achieves a 24.1% reduction in complexity and without a reduction in accuracy compared to the baseline. Through comprehensive scenario analysis, the results demonstrate that the fine-tuned RF decision-making function exhibits 99.36% spoofing detection accuracy, albeit with reduced structural and temporal complexity compared to MLP NN and FL decision-making functions. Additionally, the MLP NN decision function offers increased structural and temporal complexity at the expense of 98.12% spoofing detection accuracy compared to fine-tuned RF and FL decision-making models. Considering the trade-off between structural-temporal complexity and detection accuracy, the fine-tuned RF decision function emerges as a preferable option over MLP NN and FL.</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

Robust GNSS spoofing detector using optimized machine learning model on embedded platforms

  • 期刊论文

  • 发表日期：

    2025年11月13日

查看全文
  • K. Zarrinnegar,
  • J. Sormayli,
  • M. R. Mosavi

摘要

Global Navigation Satellite Systems (GNSS) play a crucial role in contemporary life, enabling various applications such as autonomous vehicles, agricultural industries, marine navigation, and military operations. However, the performance of GNSS receivers is susceptible to both intentional and unintentional interferences, with satellite spoofing attacks posing a significant threat. This work presents a GNSS detector receiver implemented using an STM32H7 series ARM microcontroller and a U-blox ZED-F9T receiver. Leveraging decision-making methods such as Fuzzy Logic (FL), Multi-Layer Perceptron Neural Network (MLP NN), and Random Forest (RF), as proposed in recent literature, the detector receiver aims to effectively detect spoofing attacks. In addition, the Constrained Complexity Minimization (CCM) algorithm was applied to compromise the complexity and accuracy of the RF Machine Learning (ML) model. According to the results, with the fine-tuned configuration, the optimized RF model achieves a 24.1% reduction in complexity and without a reduction in accuracy compared to the baseline. Through comprehensive scenario analysis, the results demonstrate that the fine-tuned RF decision-making function exhibits 99.36% spoofing detection accuracy, albeit with reduced structural and temporal complexity compared to MLP NN and FL decision-making functions. Additionally, the MLP NN decision function offers increased structural and temporal complexity at the expense of 98.12% spoofing detection accuracy compared to fine-tuned RF and FL decision-making models. Considering the trade-off between structural-temporal complexity and detection accuracy, the fine-tuned RF decision function emerges as a preferable option over MLP NN and FL.