From Low-Cost Spectrum Monitoring to 5G Networks: Algorithms and Systems for Localizing and Identifying Wireless Transmissions

Abstract

Electromagnetic Radiation is the effect that Electromagnetic fields can exert at a distance from their source. Depending on the frequency of this radiation, we can divide it into categories such as: Radio-Frequency (RF), Infrared, Visible Light, X-rays or Gamma rays. In particular, the RFspectrum is the foundation for all modern wireless communications. Our lives would be difficult to imagine should we not have access to cellular, GPS, WiFi, Bluetooth, IoT and many other technologies. Although the RF spectrum is heavily regulated, current advances and the availability of hardware make it easier than ever to disrupt legitimate communications. In fact, at the beginning of the decade 2020, attacks on the spectrum are becoming more and more common. Currently, spectrum regulators must deploy bulky and expensive hardware to monitor these threats, which does not scale due to the amount of manual labor required.

Software-Defined Radio (SDR) devices are a flexible alternative for monitoring the RF spectrum due to how easy it is to reconfigure their measuring parameters like sampling rate or center frequency in real-time. This flexibility has allowed researchers and engineers to apply them in numerous applications. It is possible to perform low-level tasks like signal identification, localization or even self-positioning, up to building car tracking networks via tire transmissions or 5G standard compliant core networks. Not all SDR receivers can perform all tasks, as lower cost ones tend to have limitations in their hardware, but each category of device can enable a wide range of applications nonetheless.

In this thesis, we present four applications for localizing and identifying wireless transmissions that can be achieved taking advantage of the flexibility offered by SDR receivers. The first three demonstrate that with low-cost receivers it is possible to perform a broad range of activities. The first application is a transmitter localization system that can be deployed at the city level and achieves notable accuracy without the need for additional hardware for synchronization. The second application shows that it is possible to build a self-positioning architecture that employs aircraft broadcast signals and the metadata contained within. We show that by careful modeling of the hardware and precise selection of signals, it is possible to provide results comparable to GPS or other commercial solutions, without the need for it. The last of this set explores a higher-level application, where we build a car tracking infrastructure based on capturing the signals transmitted by the pressure sensors installed on wheels. We then show that with careful processing it is possible to robustly track cars over long periods of time and infer driver patterns from this tracking. Our results not only show the feasibility of building monitoring networks with low-cost receivers but also highlight the urge to redesign less privacy-aware protocols in this and many other areas.

The fourth application makes use of higher-end SDR receivers that support better synchronization hardware to build a custom 5G compliant network. With this network, we are able to explore the localization performance that these higher-end platforms can offer, after doing several modification to the underlying platform. Our extensive experiments with different geometric configurations, number of sensors, optimization routines and processing techniques show that it is possible to achieve highly accurate positions.

This thesis makes several contributions to the proposed architectures for all applications, as well as the techniques employed. In order to validate our proposals, we perform an extensive analysis with data coming from real-world scenarios.