Signal Intelligence Analyzing Neural Network
SIANN (Signal Intelligence Analyzing Neural Network) is a multi-task learning model designed to be used with Software Defined Radios for SIGINT usage.
- Support for radio frequency signal analysis
- Input: raw I/Q samples (complex time-domain signals)
- Evaluation on real-world and synthetic datasets
- SIANN-C (Signal Intelligence Analyzing Neural Network, Classification model): Multi-task learning architecture for joint signal classification and fingerprinting
[More to be added].
- Python 3.11.13
- TensorFlow 2.15.0
- Libraries: numpy, h5py, matplotlib, pyadii-io, seaborn, scikit-learn, scipy
- GNU Radio (optional, for visualization)
- HDF5 Support
Initial training datasets were provided for use by ANDRO Computational Solutions. The following datasets were used for training the model:
- RadComAWGN:
- RadComDynamic:
- RadComOta2.45GHz:
- SDR: ADALM PlutoSDR
- Signal and Sample Processing: Raghav Sureshbabu
- Model Architecture and Development: Shane Ganz and Corey Mack
- Live Classification Model Development: Sarah Morar and Shane Ganz
- Documentation: Shane Ganz, Corey Mack, Sarah Morar, Raghav Sureshbabu
The development team would like to thank ANDRO Computational Solutions for the initial datasets used for training. The team would also like to thank Sean Furman and Dr. Sabarish Krishna Moorthy for their guidance in developing the initial model.
This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License.
If you use this work in academic research, please cite:
@misc{SIANN,
author = {Shane Ganz and Corey Mack and Sarah Morar and Raghav Sureshbabu},
title = {SIANN: Signal Intelligence Analysis Neural Network},
year = {2025},
howpublished = {\url{https://github.com/sarahmorar/SIANN}}
}
You are free to use, share, and modify this code for non-commercial purposes only.
Commercial use is strictly prohibited without prior written permission.
To request commercial licensing or exceptions, please contact: [morarsa.udm@gmail.com]