CSI-Based Cross-Technology Interference Detection and Classification on Wi-Fi 6 IoT Hardware

This repository contains a Convolutional Neural Network (CNN) trained to identify Cross-Technology Interference (CTI) based on the Channel State Information (CSI) obtained from Wi-Fi 6 devices, in particular the ESP32-C6.

There are two different folders:

• esp32c6 contains the ESP-IDF project to run it on an ESP32-C6 device, which is partly automatically generated and partly based on the iPerf example.
• model contains the Python scripts and outputs to train and test the model using PyTorch.

The trained model is available as a serialized PyTorch state dictionary (model/CSImodel_snr14_24_sir1_15.pth) and as optimized model in ONNX format (model/CSImodel_opt.onnx).

Follow this guide to build the ESP-IDF project and flash it to an ESP32-C6 device.

For access to the prepared data, please contact me at Thijs.Havinga@UGent.be. If you want to retrain and deploy the model yourself, execute model/train.py (which may take several hours). Then follow this guide to download the ESP-DL library first, and then use model/onnx_convert.py to generate the ESP-IDF project.