Content area
Global Navigation Satellite Systems (GNSS) receivers are susceptible to anthropogenic Radio Frequency Interference (RFI) attacks in the forms of jamming and spoofing due to the low power, unencrypted signals they receive from the satellite constellations. While jamming is a practice that denies a GNSS receiver the ability to form a correct Position-Velocity-Time (PVT) solution, spoofing represents a more insidious threat by using forged satellite signals that cause the victim receiver to compute a false PVT solution. Recent works have demonstrated the vulnerability of GNSS-enabled Android smartphones to common spoofing methods. Spoofing detection techniques specific to Android devices is an active area of research, and has been facilitated over the last 10 years by the GNSSMeasurement Application Programming Interface (API) which makes deep introspection into the GNSS chip onboard the device available. Nevertheless, Android smartphones present unique challenges that make comprehensive spoofing detection difficult. Chief among these challenges are the wide variety of implementations of GNSS processing at both the hardware and software levels, making it challenging to generalize the applicability of detection methods across different device models. Furthermore, limitations in the hardware available on smartphones and the restriction of data offered by the GNSSMeasurement API to the post-correlation level preclude the use of many established spoofing detection techniques developed outside of the smartphone domain. This thesis expands on previous works that utilize the GNSSMeasurement API to detect spoofing events. To accomplish this we designed and conducted experiments subjecting several different smartphones to an RF environment with both authentic and spoofed satellite signals. We collected the raw measurements from each smartphone and used the data to inform the development of a novel approach that improves spoofing detection in two key ways. First, we utilize measurement fields and derived metrics offering unique insight in the detection of spoofing previously unexplored for this purpose. Second, we use the full feature set to maximize the joint estimation potential offered by all the information available through the use of a Support Vector Machine. We demonstrate improvements in spoofing detection over existing methods, particularly in the applicability of the approach across different device models even when device-specific training data is not available.