A random forest based spoofing detection method for GNSS timing

GNSS has been widely adopted as a primary source of high-precision timing for critical infrastructures such as power grid, communications and financial systems. Its security is therefore essential to the reliable operation of these services. To address the vulnerability of GNSS timing equipment to spoofing attacks, this paper designs a random forest based spoofing detection method for GNSS timing. By leveraging the consistency in signal statistical characteristics arising from the periodic nature of GNSS satellite orbits, the method constructs high discriminative features inspired by kernel function concepts and enhance feature cooperativity through a random forest model, thereby improving detection performance. Tests with real collected data on BeiDou Navigation Satellite System (BDS) B1I and B3I signals show that with 30 decision trees, the model’s out-of-bag (OOB) error is approximately 0.13%. The F1-score, which comprehensively balances precision and recall, exceeds 99% on both the independent test set and an additional validation scenario. These results fully validate the method's high accuracy, strong robustness and good generalization capability, confirming its effectiveness in enhancing the security of GNSS timing equipment.