Towards Simple Machine Learning Baselines for GNSS RFI Detection

Machine learning research in GNSS radio frequency interference (RFI) detection often lacks a proper justification for the decisions made in deep learning-based model architectures. Our paper challenges the status quo in machine learning approaches for GNSS RFI detection, revealing the potentially misleading track of current research and highlighting alternative directions. Our position advocates for a shift in focus from solely pursuing novel model designs to critically evaluating the utility of complex black box deep learning methods against simpler and more interpretable machine learning baselines. Our findings demonstrate the need for the creation of simple baselines and suggest the need for more exploration and development of simple and interpretable machine learning methods for the detection of GNSS RFIs. The increment of model complexity in the state-of-the-art deep learning-based models often provides very little improvement. Thanks to a unique dataset from Swiss Air Force and Swiss Air-Rescue (Rega), preprocessed by Swiss Air Navigation Services Ltd. (Skyguide), we demonstrate the effectiveness of a simple machine learning baseline for GNSS RFI detection on real-world large-scale aircraft data containing flight recordings impacted by real jamming. The experimental results indicate that our solution successfully detects potential GNSS RFI with 91% accuracy outperforming state-of-the-art deep learning architectures. We believe that our work offers insights and suggestions for the field to move forward.

@article{ivanov2025_2504.07993,
  title={ Towards Simple Machine Learning Baselines for GNSS RFI Detection },
  author={ Viktor Ivanov and Richard C. Wilson and Maurizio Scaramuzza },
  journal={arXiv preprint arXiv:2504.07993},
  year={ 2025 }
}