T-PRIME: Transformer-based Protocol Identification for Machine-learning at the Edge

TL;DR

T-PRIME introduces a Transformer-based protocol classifier for real-time, edge-based identification of wireless protocols directly from raw IQ samples, addressing the limitations of correlation-based preamble matching in low-SNR and overlapping scenarios. The design uses an encoder-only Transformer that processes IQ sequences without spectral pre-processing, achieving superior accuracy over legacy methods and state-of-the-art ML baselines. The authors validate performance on synthetic data and a 66 GB OTA WiFi dataset, and demonstrate near real-time inference on the AIR-T SDR platform with a modular three-thread pipeline and TensorRT optimization. The work provides extensive OTA data, real-time deployment insights, and public code, highlighting the practicality and potential impact for spectrum sharing, interference management, and security in edge wireless systems.

Abstract

Spectrum sharing allows different protocols of the same standard (e.g., 802.11 family) or different standards (e.g., LTE and DVB) to coexist in overlapping frequency bands. As this paradigm continues to spread, wireless systems must also evolve to identify active transmitters and unauthorized waveforms in real time under intentional distortion of preambles, extremely low signal-to-noise ratios and challenging channel conditions. We overcome limitations of correlation-based preamble matching methods in such conditions through the design of T-PRIME: a Transformer-based machine learning approach. T-PRIME learns the structural design of transmitted frames through its attention mechanism, looking at sequence patterns that go beyond the preamble alone. The paper makes three contributions: First, it compares Transformer models and demonstrates their superiority over traditional methods and state-of-the-art neural networks. Second, it rigorously analyzes T-PRIME's real-time feasibility on DeepWave's AIR-T platform. Third, it utilizes an extensive 66 GB dataset of over-the-air (OTA) WiFi transmissions for training, which is released along with the code for community use. Results reveal nearly perfect (i.e. $>98\%$) classification accuracy under simulated scenarios, showing $100\%$ detection improvement over legacy methods in low SNR ranges, $97\%$ classification accuracy for OTA single-protocol transmissions and up to $75\%$ double-protocol classification accuracy in interference scenarios.

Figures (12)

• Figure 1: T-PRIME system overview and receiver design.
• Figure 2: Transformer-based model architecture.
• Figure 3: Transmission split into sequences, slices and, tokens.
• Figure 4: (a) Comparison between SM, LG Transformer-based architectures and state-of-the-art signal classification models tested on all channel models and different SNR conditions. The LG Transformer achieves the best overall accuracy and it retains high accuracy (i.e. $> 98\%$) for SNR as low as -10 dB. All models improve under optimal hyperparameter search (see Appendix), but the LG Transformer still outperforms all other methods. (b) Performance comparison for each individual channel model tested during simulation.
• Figure 5: T-PRIME running real-time classification in an AIR-T receiver. 802.11g protocol is shown to be correctly detected in this instance.