Supervisors

Suitable for

Abstract

Existing work has shown that satellite transmitters can be authenticated by looking only at the physical layer signal, due to small differences in transmitter hardware. This is particularly useful in the absence of other authentication, and
can also be used to identify transmitter hardware, classify the nature of attacks, or prevent timing-based attacks.

In this project, you will extend existing fingerprint-based systems in one of several ways:
• Through our collaboration with the European Space Agency (ESA), you will adapt this system to run on their new “CyberCUBE” satellite, looking at the uplink instead of the downlink to authenticate ground systems,
identify the transmitter hardware, or understand when the system is under attack. This will use finetuning, quantisation, and/or distillation to produce smaller or more performant models that can run under more constrained
conditions.
• Continuing to focus on the satellite downlink, you will develop techniques to improve performance. This may involve making use of more incoming data from the satellite, extracting additional features, or combining with
other systems.
• Looking beyond authentication, you will investigate novel applications for transmitter fingerprints, extracting location and other information from physical layer characteristics.

Interested students should have experience working with Python, with machine learning / tensorflow experience preferable. No prior experience in radio communication is required.

Relevant reading material:
• “SatIQ” satellite fingerprinting paper: https://www.cs.ox.ac.uk/files/14805/main.pdf
• SatIQ source code: https://github.com/ssloxford/SatIQ
• Extensions of SatIQ:
– https://arxiv.org/pdf/2402.05042
– https://arxiv.org/pdf/2503.02118