frclba/why-fi

Position tracking software based on wi-fi signals

why-fi

This is a position tracking software based on wi-fi signals. It's alpha quality, meaning it's not fancy or smooth by any means. If you intend to use this, be prepared to get your hands dirty.

You can also take a look at the original proof of concept here. It's a simple Python code that takes wi-fi packets and extracts RSSI data (signal strength) from each one. Everything else on this project builds up on top of the following premise: if we can measure the signal strength of a device in relation to a receiver, we can map its position over time.

Why have I done this?

This project shows how easy it is to track people's movements without their consent using their own smartphones.

The constant search for a Wi-Fi turn our phones into beacons signaling where we are, who we are and even where we have been. With that, it's possible to create a map of people's movements by collecting data in large scale.

Despite looking like a complex endeavour, doing something like this is fairly simple, so I created an open source version of such tracking system to show that any small organization (a shopping center, for instance) can track us and, with this, bring awareness to this issue.

I have presented this work at the Mozilla Festival, in London, in 2019. You can get the slides here.

How to use this project

The work will involve three steps:

1. collect the wi-fi packets
2. filter and process the data to turn it into a table of movements
3. visualize the movements in a web browser

To do this, you need to have a working knowledge of:

• Linux
• SSH
• Python/Jupyter
• Javascript
• Docker (not required, but recommended)

How to capture packets

Until now, I have used Raspberry Pi 3B with great results. Certainly, you can use other Raspberry Pi models, or even completely different hardware for this step. All you really need is a device capable of running airodump-ng.

Raspberry Pi 3B

There is a helper script that you can use to ease the data collection process. In essence, its just a way to juggle airodump-ng functionality and the regular wi-fi functionality. You won't need it if you can connect to the Pi using an ethernet cable or keyboard/screen during the data gathering process.

First, download the "Kali Linux Raspberry Pi 2 (v1.2), 3 and 4 64-Bit" from here.

Now, flash it to a SDCard and boot it in your Raspberry Pi 3B. If you need help with that, take a look here.

Copy the helper script to Kali's home directory and run:

cd ~
chmod +x packet_collect.sh
sudo ./packet_collect.sh

Since the airodump will use the wi-fi hardware, any connection through it will drop. The captured packets file name will have the following format 'rasp-<local_mac_addr>-<job_start_time>.cap'.

The script will blink the Pi's leds to let you know if everything is working as intended. The "color code" is as follows

• red and green blinking separately: everything is working fine and the data is being collected
• red and green blinking together: something went wrong with airodump, but the helper script took notice and restored the wi-fi functionality, so you can use SSH to figure out what went wrong
• red and green on together (no blinking): the packet collection process went fine and the script exited normally; wi-fi functionality is restored and you can use SSH to get the file

Any combination different from the above probably means that the data collection process has gone wrong and the helper script have failed to notice it. In this case, reboot the Pi to recover network functionality (or plug an ethernet cable).

Process the packets

Enter the lab folder on this project. Run Jupyter with the following Docker command (if you don't have Docker, follow the instructions here):

docker-compose up

Put all the .cap files in the captured_packets folder and run the Packets_pipeline.ipynb notebook. It can take a couple of minutes to run depending of the amount of data you have. When its done, you will have a movements.csv file in the lab folder.

Visualize the data

Go to the frontend folder, place the movements.csv on it and run:

docker-compose up

Access http://localhost to see a graph visualization of the detected movements. The final result will probably look something like this:

---

Big thanks to everyone that made this project possible. But specially to @tiagofassoni for the original idea.