Pwnagotchi 64-Bit AI Edition
For an in-depth guide — how it works, everything this fork changes versus upstream, and how to use
it — see USER_GUIDE.md.
What is Pwnagotchi?
Pwnagotchi is a WiFi-auditing tool that carries a face and a personality, running on a small piece
of hardware (usually a Raspberry Pi) you take with you. The actual attack work — deauthing,
associating, capturing handshakes — is all done by bettercap; a
lot of people assume the “AI” is what’s hacking the WiFi, but it isn’t. What the AI does is learn
how to run bettercap well: how long to linger on a channel before hopping, how aggressively to
hop between channels, how long to wait for a handshake before giving up on a target, how weak a
signal is even worth bothering with, and a dozen other timing/behavior knobs. Every epoch, it sees
what those settings produced — handshakes captured, time spent blind, targets missed — and that
becomes a reward signal nudging its next set of settings. Over time it learns which behavior
actually works in whatever environment it’s really operating in, instead of running forever on one
fixed configuration that might be great sitting still on a desk and useless out on a walk.
Credit for the original concept and the reinforcement-learning core goes to
evilsocket, who created Pwnagotchi — see Acknowledgments below.
About This Fork
This build keeps that same reinforcement-learning core but rebuilds the platform underneath it:
- 64-bit, PyTorch/stable-baselines3 instead of the original’s older 32-bit, TensorFlow-based
A2C implementation — faster epoch processing and better training stability.
- Kali Linux base instead of the legacy 32-bit image, for native Nexmon firmware support —
reliable monitor mode and packet injection without extra driver work.
Hardware Support: specifically optimized for the Raspberry Pi Zero 2 W with waveshare
eink 2.13 v4
*Note: this image may run on other ARM64 devices with different setups.
Configuration / Usage
ssh login is user: pwn , password: raspberry
default ip for gadgetmode is 10.42.0.2
User configuration file is at /etc/pwnagotchi/config.toml.
Created/edited by running sudo pwnagotchi --setup.
Do not edit /etc/pwnagotchi/default.toml — it is overwritten on every restart.
Building from Source
This project uses Docker to create a clean, reproducible build environment. This ensures your system stays clean and the build succeeds regardless of your local Linux distribution.
Requirements:
- Docker installed and configured.
Instructions:
- Clone the repository.
- Run:
make
The build process will automatically:
- Package the source code.
- Launch an isolated Debian container.
- Download the official Kali base image and apply all security patches, Bluetooth drivers, and custom UI plugins.
- Output the final, ready-to-flash image to the
pwnagotchi64/dist/ folder called pwnagotchi64-0.0.0.0.img
Flashing SD Card
The easiest way is to use Raspberry PI Imager.
dont use any pre-setup features like wifi, you will break it.
Acknowledgments & Credits
This 64-bit build would not be possible without the foundational work and continuous community efforts of the following developers. My fork builds directly upon their heavy lifting:
- evilsocket - The original creator and architect of the Pwnagotchi project.
- aluminum-ice - For crucial contributions to the core codebase.
- jayofelony - For crucial contributions to the core codebase.
If I have inadvertently used your code, script, or concept without proper attribution, thank you for your indirect help! Please open an issue so I can ensure you are properly credited here.
This project is open-source and inherits the original GPL-3.0 License.
评论