Priest, Eli (2021) Unique Identification of Bluetooth Transmitters Through RF Fingerprinting. [USQ Project]
|
Text (Project)
PRIEST Eli dissertation_redacted.pdf Download (2MB) | Preview |
Abstract
Physical layer radio frequency (RF) fingerprinting has been used in military and civilian applications to identify RF transmitters for spectrum management purposes, and has been considered as a mechanism to improve assurance that a transmitter is not an impostor. It relies on the presence of observable device-specific variations to expected signal output, even between transmitters of the same type. Minor fluctuation in component values during transmitter assembly–and even placement of those components–can result in minor variances to frequency synthesis systems, modulator subsystems, and RF amplifiers, all which can be observed and used to characterise the transmitter. The complexity of the variations makes these characteristics inherently difficult to reproduce, and technically difficult to obscure.
RF fingerprinting of Bluetooth devices has been explored in the literature, but there is not sufficient information to reproduce the transient extraction stage used to produce the high-results of others. Additionally, there has been little reported work on the effects of expected environmental variables (temperature, motion, low signal to noise ratio) on classification success. This dissertation expands the existing literature by investigating the implementation and performance of a physical layer RF fingerprinting system, and the effect of real-world environmental conditions on system performance.
A downconverter was constructed to shift the entire Bluetooth band (2400–2480 MHz) down to 20–100 MHz, allowing acquisition of the entire band with low-cost acquisition hardware (i.e. a PicoScope 5444B). An RF fingerprinting system, specifically the transient detection sub-system and feature extraction sub-system, is implemented in MATLAB®. Energy Criterion is confirmed as an excellent method for detecting the start of a transient portion. Additionally, a new method for detecting the end of the transient is introduced, based on the settling time of the envelope. These two methods successfully extracted the transients from several waveforms reliably; however, some transmitter types were observed to produce waveforms with significant ripple to the steady-state envelope, causing unreliable operation of the transient detection system.
To support classification a feature extraction system was implemented in MATLAB®. Features are extracted from the energy envelope and the time-frequency-energy distribution (TFED) of the signal. A link is identified between inconsistent transient length detection and inconsistent features. Classifiers were implemented using MATLAB®’s Classification Learner app, with the optimum classifier found to be a Support Vector Machine, which confirms existing literature.
A new dataset of turn-on transients was acquired for 17 devices using the constructed downconverter and acquisition system. This dataset, and an existing reference dataset, were used to assess the transient detection, feature extraction, and classifier sub-systems, and the results compared. After optimisation, classifiers were able to correctly attribute waveforms from the reference dataset to a specific device with an accuracy of 32.6%, while correct attribution when using the acquired dataset was 69.9%. When the classifiers were used to attribute waveforms to a device-type, as opposed to specific device, prediction success increased to 92.6%. This research was unable to reproduce the extremely high results (over 99% success) reported in the literature. Further work in the field, specifically improvement to the transient detection stage, is required to make RF fingerprint classification of Bluetooth devices more viable.
|
Statistics for this ePrint Item |
| Item Type: | USQ Project |
|---|---|
| Item Status: | Live Archive |
| Faculty/School / Institute/Centre: | Historic - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering (1 Jul 2013 - 31 Dec 2021) |
| Supervisors: | Leis, John |
| Qualification: | Bachelor of Engineering (Electrical and Electronic) |
| Date Deposited: | 03 Jan 2023 03:50 |
| Last Modified: | 26 Jun 2023 01:42 |
| Uncontrolled Keywords: | radio frequency, RF, fingerprinting, physical layer radio frequency fingerprinting, transmitter, bluetooth, environment, classification, |
| URI: | https://sear.unisq.edu.au/id/eprint/51828 |
Actions (login required)
 |
Archive Repository Staff Only |
