Localizing Spectrum Offenders Using Crowdsourcing

Localizing transmitters using crowdsourced data is considered a practical and cost-effective method, enabling real-time and accurate location tracking over large areas in order to provide protection to spectrum users. This chapter explores various recent RSS-based localization techniques which use crowdsourced measurements, including path loss models, fingerprinting, and machine learning-based approaches. Our focus is on utilizing convolutional neural networks to improve localization accuracy for both single and multiple simultaneous transmitters. We also delve into the vulnerability of a crowdsourced system to adversarial attacks and present a case study demonstrating successful attacks on a real world dataset. We show the effectiveness of adversarial training as a defense mechanism, and highlight challenges in developing practical localization systems.