The UAV Signal (UAVSig) dataset was collected to develop drone and remote controller identification using RF fingerprinting. This dataset includes captures from multiple drones and controllers, with more detail about the experimental setup found in our paper [1] and in the information presentation found in the dataset documentation.

Important Notes:

• The dataverse interface is unstable sometimes. If the page fails to load, you can try opening it during different time and/or with different networks.
• It may overload the server if you try to download all files using the web interface as one zip file. As a workaround, we have a Python script (dataverse_load.py) to automate the data file downloading. Please open and examine the important arguments in the script. Be sure to modify the version number to reflect your desired dataset version number. The latest version is listed on the Dataverse page just below the dataset name.
• Also, we have found a limitation of the dataset collection setup: some received sample segments were dropped by the B205mini SDR, so there are random gaps in the posted raw IQ data. The labels still properly reflect the received signals, and the listed capture parameters are accurate to our OTA setup. We are currently developing an alternative dataset with a wider capture bandwidth and without dropping samples.

References:
[1] T. Zhao, B. W. Domae, C. Steigerwald, L. B. Paradis, T. Chabuk, D. Cabric, “Drone RF Signal Detection and Fingerprinting: UAVSig Dataset and Deep Learning Approach,” accepted to the IEEE Military Communications Conference (MILCOM) 2024