Thesis Defence: An Unsupervised Machine Learning Approach to Real-Time Spectrum Sensing in Dynamic Communication Channels

Eli Garlick, supervised by Dr. Md Jahangir Hossain, will defend their thesis titled “An Unsupervised Machine Learning Approach to Real-Time Spectrum Sensing in Dynamic Communication Channels” in partial fulfillment of the requirements for the degree of Master of Applied Science in Electrical Engineering.

An abstract for Eli Garlick’s thesis is included below.

Defences are open to all members of the campus community as well as the general public. Please email jahangir.hossain@ubc.ca to receive the Zoom link for this defence.

---

ABSTRACT

Cognitive tactical wireless networks (TWNs) require spectrum awareness to avoid interference and jamming in the communication channel and ensure quality of service in data transmission. Conventional supervised machine learning (ML) algorithms’ capability to provide spectrum awareness is challenged by the requirement of labelled interference signals. Due to the dynamic nature of interference signals in the frequency bands used by cognitive TWNs, it is difficult to manually acquire labelled datasets for all interference signals. Detecting the presence of an unknown and remote interference source in a frequency band from the transmitter end is also challenging, especially when the received interference power remains at or below the noise floor. To address these issues, this thesis proposes an automated interference detection framework, called MARSS (Machine Learning Aided Resilient Spectrum Surveillance). MARSS is a fully unsupervised method, which first extracts low-dimensional representative features from spectrograms by suppressing noise and background information and using a Convolutional Neural Network (CNN) with a novel loss function. It then distinguishes signals with and without interference by applying an isolation forest model on the extracted features. The uniqueness of MARSS is its ability to detect hidden and unknown interference signals in multiple frequency bands without using any prior labels, thanks to its superior feature extraction capability. The capability of MARSS is further extended to infer the level of interference by designing a multi-level interference classification framework. Using extensive simulations in GNURadio, the superiority of MARSS in detecting interference over existing ML methods is demonstrated. The effectiveness of MARSS is also validated by over-the-air (OTA) experiments using software-defined radios.