Edge AI LoRa Jamming Detection

Massive GPS jamming campaigns across Europe have proven that physical-layer electronic warfare is no longer just a military problem—it is a direct threat to civil infrastructure. Off-grid mesh networks like Meshtastic are the ultimate backup lifeline for private communication, but because they operate on open public frequencies, they are sitting ducks. Anyone with a cheap, pocket-sized transmitter can easily flood the airwaves and completely silence an entire local network.

I built this project to turn these nodes into smart, self-defending units. Instead of relying on heavy external gear or cloud analysis, this firmware packs an optimized Neural Network directly into the device's local memory. It reads the signals in real-time, instantly identifying and alerting on malicious interference. It is tactical electronic defense scaled down to a microchip, ensuring real time information about the quality of private communications and allows for corrections to avoid jamming.

To achieve this, the firmware embeds an ultra-optimized Edge AI Neural Network directly onto the node's local microcontroller. It runs real-time Spectral Power FFT analysis on the airwaves, achieving a 1ms inference time using a mere 1.4KB of peak RAM so the main network routing remains entirely unhindered. To handle the chaos of real-world hardware, I coupled the AI with an AGC-resilient tracking algorithm and a signal hold timer. This stops sudden hardware gain shifts from triggering false alarms and stabilizes the screen during high-speed pulsed attacks—delivering tactical electronic defense on a single microchip.

more info and access to the AI model training data on my website:

https://jalbertomoro.netlify.app/?project=rf-detector

There is a web flasher available in the repository which allows the firmware to be installed directly from the browser to the device using only a usb-c cable. ( hold down the boot button before connecting the cable in order to be able to reprogram the board).

The code in the repository can be compiled for the device using the platformio extension in VS Code.