RF Signals Detection, Analysis & Clustering with ML to a Specific Air Object

Machine-Learning-based ADS-B Signal Analysis

This project performs RF signal detection and classification from aircraft ADS-B transmissions
using a Software Defined Radio (RTL-SDR) and the "dump1090" decoder.

The device which been used is the "RTL2832U TV Tuner for Laptop / PC with DVB-T Terrestrial Receiver and USB-A connection".

Chip	Role
RTL2832U	digital interface (USB → PC) & baseband processor
Tuner chip (usually R820T / R820T2)	adjusts the reception frequency (24 MHz – 1766 MHz)

The R820T2 is acting like the RF tuner: selects the frequency (e.g. 1090 MHz for ADS-B), demodulates => and sends the I/Q samples to the RTL2832U. The conversion that I did in practice. I didn't change the hardware — I unlock its capabilities through driver software that makes it work like an SDR.

For the Windows system:

1. Install Zadig
2. Replace the “Bulk-In Interface (Interface 0)” driver with the WinUSB driver With this way the operating system sees the tuner as an "RTL-SDR dongle" instead of a "TV tuner"!

For the macOS / Linux:

1. Install the rtl-sdr package
2. The librtlsdr driver is talking directly to the chip (bypassing the DVB-T firmware)
3. The dump1090 program is using this I/Q data to decode ADS-B packets

How it became an "RF receiver"?

With this process, the tuner no longer searches for TV channels and in parallel is open to any RF frequency within its spectrum (e.g. 24 MHz – 1766 MHz). After this dump1090 takes the raw RF samples Decodes the Mode-S / ADS-B packets (1090 MHz) and sends the decoded data (ICAO, altitude, lat/lon, speed, etc.) to port 30003 → where your Python logger is connected.

What does it mean in practice? A 25€ TV tuner becomes an SDR receiver for RF signals (24 MHz – 1.7 GHz) which is Able to detect:

1. aircraft (ADS-B at 1090 MHz)
2. AIS from ships (162 MHz)
3. FM radio (88–108 MHz)
4. NOAA weather satellites (~137 MHz)
5. GPS L1 (1.575 GHz) even RF noise/interference with FFT/Waterfall

USAGE from terminal:

1. Start dump1090 server dump1090 --interactive --net
2. Run the logger python adsb_logger_enhanced.py
3. Analyze saved data python adsb_analyzer_ML.py

Compliance & Safety — Passive Reception Only

Attestation: This project performs passive RF reception only. I do not transmit or broadcast any RF signals as part of this work.

What we do

• Use an inexpensive consumer RTL-SDR USB tuner (RTL2832U + R820T/R820T2) configured as an SDR receiver.
• Decode ADS-B (1090 MHz) packets using dump1090 and process them offline or stream decoded messages over a local TCP socket for analysis.
• All software and scripts in this repository perform decoding and analysis only — there is no transmission code.

Hardware / Driver notes

• Device used: RTL2832U tuner (DVB-T USB dongle). No hardware modifications were made.
• On Windows the device driver was replaced using Zadig with WinUSB so the dongle acts as an RTL-SDR receiver only.
• On macOS/Linux the rtl-sdr library (librtlsdr) is used to access I/Q samples; this library supports reception only for this device.
• We do not install or use transmit-capable drivers/software (e.g., no HackRF tools, no bladeRF host programs, no soapy transmit modules).

Legal & safety reminder

• Regulations differ by country. This repository documents passive reception only. If you intend to transmit, check and obtain the required licenses/permits from your national communications regulator before doing so.
• For questions on local law, contact your national authority (e.g., FCC in the USA, Ofcom in the UK, EETT/RAE in Greece).