Date of Award
2026
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Computer Science
Committee Chair
Tathagata Mukherjee
Committee Member
Feng Zhu
Committee Member
Jacob Hauenstein
Committee Member
Aleksandar Milenkovic
Committee Member
Sudhir Aggarwal
Research Advisor
Tathagata Mukherjee
Subject(s)
Wireless communication systems--Security measures, Radio frequency--Identification, Deep learning (Machine learning)
Abstract
This dissertation introduces a deep learning based implementation of a multitask learning framework called the Common and Specialized Learning (CSL) for applications to problems in the Radio Frequency (RF) domain. We demonstrate the use of the framework for two simultaneous multi-classification problems in the RF domain. CSL employs a shared common learning module to extract high-level features, followed by specialized task-specific modules to refine these common features for downstream classification tasks. This two-step structure of the learner reduces optimization time and regularizes the training using shared general features. Furthermore, for the first problem discussed in this dissertation, a novel pre-processing technique called the Generalized Multi-Angular Projection (GMAP) is introduced to enhance the raw in-phase and quadrature (I/Q) RF data by generating multi-perspective projection for improving feature discrimination for RF identification tasks. The efficacy of the CSL framework is validated through two applications in the RF domain. For both these applications, we built an end-to-end implementation that operates directly on the raw I/Q signal data or its projection using GMAP, and performs multi-classification. The first problem addresses the simultaneous identification of the transmitter, modulation, and receiver (TMR) in uncharacterized RF environments, while the second addresses the problem of estimating receiver position and orientation using FM signals in uncharacterized environments. Both applications can be used to improve the security of RF systems.
Recommended Citation
Sulaiman, Moath Mustafa MJ, "Common and specialized learning with applications to radio frequency security" (2026). Dissertations. 506.
https://louis.uah.edu/uah-dissertations/506