A Deep Reinforcement Learning Framework for GNSS Spoofing Detection and Classification

Satya S. Vemuri, Jari Nurmi, M. Zahidul H. Bhuiyan, Saiful Islam, Elena Simona Lohan, Joaquin Torres Sospedra

Peer Reviewed

Abstract:	Spoofing is an escalating threat to satellite navigation systems such as GNSS, requiring intelligent and adaptive detection mechanisms. This work proposes a Deep Reinforcement Learning (DRL) framework that formulates spoofing detection and classification as a Markov Decision Process (MDP). Tracking loop features—such as carrier-to-noise ratio (C/N0), earlyprompt-late correlator outputs, Delay-Locked Loop (DLL) and Phase-Locked Loop (PLL) discriminator values serve as states, while detection and classification decisions are modeled as actions. A reward function based on detection accuracy guides the learning process. Two Deep Reinforcement Learning (DRL) agents, Deep Q-Network (DQN), value-based and Proximal Policy Optimization (PPO), which is policy-based are evaluated for GNSS spoofing detection and classification. PPO demonstrates superior performance and stability across diverse spoofing scenarios, achieving an accuracy of 99%, while DQN follows with 94%. Both agents learn to exploit subtle distortions in signal tracking behavior, enabling robust spoofing identification without external authentication. GNSS data corresponding to both targeted and untargeted spoofing attacks, provided by the Finnish Geospatial Research Institute (FGI), is utilized for analysis and to demonstrate the effectiveness of DRL based detection and classification approaches.
Published in:	Proceedings of the 2026 International Technical Meeting of The Institute of Navigation January 26 - 29, 2026 Hyatt Regency Orange County Anaheim, California
Pages:	414 - 427
Cite this article:	Vemuri, Satya S., Nurmi, Jari, Bhuiyan, M. Zahidul H., Islam, Saiful, Lohan, Elena Simona, Sospedra, Joaquin Torres, "A Deep Reinforcement Learning Framework for GNSS Spoofing Detection and Classification," Proceedings of the 2026 International Technical Meeting of The Institute of Navigation, Anaheim, California, January 2026, pp. 414-427. https://doi.org/10.33012/2026.20554
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In