Session A2: GNSS Security: Interference, Jamming, and Spoofing 1

A Particle Filter Based Sensor Fusion Approach for GNSS Spoofing Detection Incorporating Ground Vehicle Navigation Constraints
Muhammad Sami Irfan, Sagar Dasgupta, and Mizanur Rahman, The University of Alabama
Location: Beacon A

We develop a pseudorange anomaly detection method for spoofing detection based on particle filter and leveraging Inertial Measurement Unit (IMU) sensor data and High-definition (HD) map. This method aims to enhance the robustness of GNSS/INS (Inertial Navigation System) integration against spoofing attacks by using multiple sources of information to detect anomalies in pseudorange measurements. During spoofing of GNSS signals, the spoofer attempts to create a shift in either the time or the positions of the receiver or both. This entails a manipulation of the pseudorange from each of the spoofed tracked satellites which is achieved by artificially manipulating the code phase delays of the incident signal at the receiver antenna. As Commercial Off-The-Shelf (COTS) receivers can report observed pseudoranges, we develop a pseudorange based method for spoofing detection which can be readily deployed in existing navigation systems with only minimal software changes. We leverage the fact that an autonomous vehicle travelling in a structured road network will have access to detailed HD map of its intended route which gives precise location of road feature of a road segment. We utilize this in conjunction with prior satellite position data, ephemeris data and receiver dynamics from IMU data to generate particles that represent the estimates for pseudoranges of the tracked satellites. Therefore, these estimates incorporate receiver movement as well as the physical constraints of its location shift between epochs. A spoofer attempting to spoof and divert the location of the receiver will have to induce deviations that violate the constraints of the vehicles motion along its intended route from its origin to its destination. Consequently, once new observations are available, we track the likelihood of the observations against the estimates. If for a sufficiently large number of particles, the current observations have low likelihood, we asses that the observations have been subjected to spoofing. The number of particles required for spoofing detection is based on an adaptive threshold which takes into account variances in the data used for estimation. The designed filter system employs a multiple filter approach whereby each of the tracked satellites are monitored by individual filters. This approach allows for independent tracking of the pseudoranges, making it easier to identify anomalies affecting specific satellites. However, the individual filter operating in isolation can be prone to divergence which may compromise detection reliability. We prevent this by utilizing a secondary filter which monitors the individual outputs of the filters for any divergence to ensure consistent operation and makes corrections to any diverging filters by adjusting the weights of the particles. This hierarchical filtering strategy ensures that the system remains stable and effective, even when subjected to complex environments. We evaluate the proposed method using real-world spoofing datasets and simulated data and compare its performance against other pseudorange-based detection techniques. Our results demonstrate that the integration of IMU, HD maps, and particle filters can enhance the spoofing detection capabilities of a receiver. The developed method presents a novel and practical solution for spoofing detection in tightly coupled GNSS/INS systems, paving the way for more secure and resilient ground navigation.