Computer Science and Engineering, Department of


First Advisor

Nirnimesh Ghose

Second Advisor

Byrav Ramamurthy

Third Advisor

Lisong Xu

Date of this Version

Fall 7-2023

Document Type



A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfilment of Requirements For the Degree of Master of Science, Major: Computer Science, Under the Supervision of Professor Nirnimesh Ghose. Lincoln, Nebraska: August, 2023

Copyright © 2023 Fahmida Afrin


Radio fingerprinting is a technique that validates wireless devices based on their unique radio frequency (RF) signals. This method is highly feasible because RF signals carry distinct hardware variations introduced during manufacturing. The security and trustworthiness of current and future wireless networks heavily rely on radio fingerprinting. In addition to identifying individual devices, it can also differentiate mission-critical targets. Despite significant efforts in the literature, existing radio fingerprinting methods require improved robustness, scalability, and resilience. This study focuses on the challenges of spatial-temporal variations in the wireless environment. Many prior approaches overlook the complex numerical structure of the in-phase and quadrature (I/Q) data by treating real and imaginary components separately. This approach results in the loss of essential information encoded in the signal's phase and amplitude, leading to lower accuracy. This thesis proposes several enhancements. First, we treat the entire complex structure of the I/Q data as a single input to a complex-valued convolutional neural network (CVNN), thereby improving the model's accuracy. Second, conduct extensive experiments to determine optimal pre-processing parameters, ensuring that over-optimistic conclusions about RF fingerprinting performance are avoided. Third, we compare various activation functions and transfer learning-based fine-tuning and a triplet network to address the variations the wireless environment introduces in scenarios involving different locations and times. We use the concept of a ``device rank'' metric to perform device identification with certainty based on RF fingerprinting. Our work concretely proves that CVNN outperforms CNN for radio fingerprinting. Concatenated Rectified Linear Units (CReLU) activation function and fine-tuning-based transfer learning perform the best for cross-location and time device fingerprinting.

Adviser: Nirnimesh Ghose