Designing sensor networks to protect primary users in spectrum access systems

In response to increasing demand on the radio frequency spectrum, regulators are implementing new spectrum access systems to share spectrum among diverse users. These systems leverage cognitive radio concepts to automatically identify suitable spectrum for users, avoiding harmful interference with higher priority users. An infrastructure of spectrum sensors is a key component of the spectrum access system, providing the means for the system to identify necessary protections. The geolocation precision of the sensing system will be limited in practice due to privacy concerns of the priority users as well as cost limitations. In this work, we examine the design options for the sensing component of a spectrum access system in terms of user performance and privacy. We apply machine learning techniques to treat the problem of estimating the priority users' state from sensor measurements, finding that the geolocation precision is closely tied to the density of the sensor deployment, and that the secondary user performance degrades rapidly if the density is too low, highlighting a tradeoff between priority user privacy and secondary user performance.