In the last few decades, Global Navigation Satellite System (GNSS) has become an indispensable element in our society. Currently, GNSS is used in a wide variety of sectors and situations, some of them offering critical services, such as transportation, telecommunications, and finances. For this reason, and combined with the relative ease an attack on the GNSS wireless signals can be performed nowadays with an Software Defined Radio (SDR) transmitter, GNSS has become more and more a target of wireless attacks of diverse nature and motivations. Nowadays, anyone can buy an interference device (also known as a jammer device) for a few euros. These devices are legal to be bought in many countries, especially online. But at the same time, they are illegal to be used. These devices can interfere with signals in specific frequency bands, used for services such as GNSS. An outage in the GNSS service at a specific location area (which can be even a few km²) could end up in disastrous consequences, such as an economical loss or even putting lives at risk, since many critical services rely on GNSS for their correct functioning.

Fundamentally, this thesis focuses on developing new methods and algorithms for interference management in GNSS. The main focus is on interference detection and classification, but discussions are also made about interference localization and mitigation. The detection and classification algorithms analyzed in this thesis are chosen from the point of view of the aviation domain, in which additional constraints (e.g., antenna placement, number of antennas, vibrations due to movement, etc.) need to be taken into account. The selected detection and classification methods are applied at the pre-correlation level, based on the raw received signal. They apply specific signal transforms in the digital domain (e.g., time-frequency transformations) to the received signal. With such algorithms, interferences can be detected at a level as low as 0 dB Jamming-to-Signal Ratio (JSR). The interference classification combines transformed signals with previously trained signals Convolutional Neural Network (CNN) and/or Support Vector Machine (SVM) to determine the type of interference signal among the studied ones. The accuracy of such a classification methodology is above 90%. Knowing which signal causes interference we can better optimize which mitigation and localization algorithm we should use to obtain the best mitigation results.

Furthermore, this thesis also studies alternative positioning methods, starting from the premise that GNSS may not always be available and/or we are certain that we can not rely on it due to some reason such as high or unmitigated interferences. Therefore, if one needs to get a Position Velocity and Time (PVT) solution, one would have to rely on alternative signals that could offer positioning features, such as the cellular network signals (i.e. 4G, 5G, and further releases) and/or satellite positioning based on Low Earth Orbit (LEO) satellites. Those systems use presumably different frequency bands, which makes it more unlikely that they will be jammed at the same time as the GNSS signal. In this sense, positioning based on LEO satellites is studied in this thesis from the point of view of feasibility and expected performance.

Cotutelle -yhteisväitöskirja