Fingerprinting radio transmitters
Radio Frequency (RF) transmitters are now ubiquitous, being found in a variety of devices such as mobile phones or military communication systems.
We can take the ‘RF fingerprint’ of a radio transmitter by quantifying incidental variations in the signal characteristics embedded in the received signal. These fingerprints originate from component variations and are thus distinct in each specific transmitter.
The goal of this project was to use Artificial Intelligence (AI) to extract this RF fingerprint, and to make a comparison against traditional methods based on statistical signal processing techniques. Convolutional Neural Networks (an AI framework often used for image processing) were trained on synthetic data. The AI approach outperformed the traditional methods across a number of performance measures, in particular speed and tolerance to random distortions in the signal. As well as civilian benefits, better RF fingerprinting techniques could improve the detection of hostile transmitters, keeping troops safer in war zones.
Securing our future
Electrical and Electronic Engineering