Blind Spoofing Detection for Multi-Antenna Snapshot Receivers using Machine-Learning Techniques
Date Issued
2020-09
Author(s)
van der Merwe, Rossouw
Nikolikj, Ana
Kram, Sebastian
Lukcin, Ivana
Nadzinski, Gorjan
Rugamer, Alexander
Felber Wolfgang
DOI
10.33012/2020.17564
Abstract
Spoofing, the transmission of false global navigation satellite system (GNSS) signals, is a problem for a GNSS receiver. Therefore,
a spoofing attack should be detected by a receiver to ensure the integrity of the position, velocity, and time (PVT) solution.
Detecting an attack is more difficult for a snapshot receiver, as temporal changes cannot be used as detection metrics. Further,
if the spoofing attacker has access to the receiver, then ideal conditions for spoofing can be facilitated. This paper presents a
machine learning (ML) approach of detecting a spoofing attack on a multi-antenna snapshot receiver. Blind detection methods
are incorporated, as it is assumed that the antenna array could have been tampered with. The ML approaches include logistic
regression (LR), K-nearest neighbors (KNN), na¨ıve Bayes (NB), decision tree (DT) and support vector machine (SVM) algorithms.
To ensure sufficient variance for training of the models, a spoofing simulation platform is developed and described in the
paper. Training and testing is done on both simulated and real world data sets. Preliminary results indicate good classification,
when training on the simulated data and validating on the real recorded data. Several of the ML methods have a classification
f1-score exceeding 99 %. Even simple ML methods, like LR, KNN and NB, show good performance, indicating that the selected
features are already adequately separating the spoofing and real data. This paper represents the first adaption of ML methods to
snapshot based spoofing detection.
a spoofing attack should be detected by a receiver to ensure the integrity of the position, velocity, and time (PVT) solution.
Detecting an attack is more difficult for a snapshot receiver, as temporal changes cannot be used as detection metrics. Further,
if the spoofing attacker has access to the receiver, then ideal conditions for spoofing can be facilitated. This paper presents a
machine learning (ML) approach of detecting a spoofing attack on a multi-antenna snapshot receiver. Blind detection methods
are incorporated, as it is assumed that the antenna array could have been tampered with. The ML approaches include logistic
regression (LR), K-nearest neighbors (KNN), na¨ıve Bayes (NB), decision tree (DT) and support vector machine (SVM) algorithms.
To ensure sufficient variance for training of the models, a spoofing simulation platform is developed and described in the
paper. Training and testing is done on both simulated and real world data sets. Preliminary results indicate good classification,
when training on the simulated data and validating on the real recorded data. Several of the ML methods have a classification
f1-score exceeding 99 %. Even simple ML methods, like LR, KNN and NB, show good performance, indicating that the selected
features are already adequately separating the spoofing and real data. This paper represents the first adaption of ML methods to
snapshot based spoofing detection.
Subjects
File(s)![Thumbnail Image]()
Loading...
Name
CONFERENCE - Blind Spoofing Detection for Multi-antenna Snapshot Receivers using Machine Learning Techniques.pdf
Size
1.78 MB
Format
Adobe PDF
Checksum
(MD5):7686f25feaa2d21d2e8dfd7eba678813