Graduate Thesis Or Dissertation
 

Leveraging Multiple Transmissions and Receptions for Channel-Agnostic Wireless Device Identification

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/0k225k019

Descriptions

Attribute NameValues
Creator
Abstract
  • Enabling accurate and automated identification of wireless devices is critical for ensuring secure and authenticated data communication in large-scale networks such as IoT networks. In the aim of devising practical identification techniques that are immune to spoofing, hardware-driven RF fingerprinting using deep neural networks, which leverages the inevitable presence of transmitter manufacturing impairments to uniquely identify devices, has recently emerged as a potential solution for such device identification problems. However, although deep neural network frameworks are proven efficient in classifying devices based on hardware impairments, such frameworks perform poorly when considering the impact of the wireless channel. That is, although when training and testing these neural networks using data generated during the same period achieve high and reliable classification accuracy, testing these networks on data generated at different times is shown to degrade the accuracy substantially due to the impact of the wireless channel, an already well recognized problem within the community. In an effort to address this challenging problem, in this thesis, we propose to leverage MIMO communication capabilities to mitigate the channel effect on the transmitted IQ symbols, thereby providing a channel-resilient device classification technique. To the best of our knowledge, we are the first to propose to leverage MIMO capabilities to mitigate the impact of channel variation on device classification. In this proposed framework, we show that combining multiple signals received over AWGN channels improves the CNN training and testing accuracy respectively by up to 20% and 30% when compared to conventional approaches. We also show that using channel estimation techniques enabled by MIMO through Space-Time Block Codes increases the robustness of deep learning based classification accuracy against Rayleigh channel variations substantially. Specifically, we show that compared to the conventional approaches, the proposed technique improves the testing accuracy by up to 40% when the learning models are trained and tested over the same channel, and by up to 60% when the models are tested on a channel that is different from that used for training.
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Rights Statement
Related Items
  • Research group website
Funding Statement (additional comments about funding)
  • This work was supported in part by the US National Science Foundation under NSF award No. 1923884.
Publisher
Peer Reviewed
Language
Embargo reason
  • Pending Publication
Embargo date range
  • 2021-08-04 to 2023-01-01

Relationships

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details BashaNora2021.pdf 2021-08-03 Public Download