Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/brannondorsey/passgan
A Deep Learning Approach for Password Guessing (https://arxiv.org/abs/1709.00440)
https://github.com/brannondorsey/passgan
deep-learning gan machine-learning password password-cracking password-strength
Last synced: 28 days ago
JSON representation
A Deep Learning Approach for Password Guessing (https://arxiv.org/abs/1709.00440)
- Host: GitHub
- URL: https://github.com/brannondorsey/passgan
- Owner: brannondorsey
- License: mit
- Created: 2017-12-20T01:35:31.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2023-02-24T07:01:06.000Z (over 1 year ago)
- Last Synced: 2024-10-14T11:01:12.922Z (28 days ago)
- Topics: deep-learning, gan, machine-learning, password, password-cracking, password-strength
- Language: Python
- Size: 26 MB
- Stars: 1,765
- Watchers: 52
- Forks: 359
- Open Issues: 24
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# PassGAN
This repository contains code for the [_PassGAN: A Deep Learning Approach for Password Guessing_](https://arxiv.org/abs/1709.00440) paper.
The model from PassGAN is taken from [_Improved Training of Wasserstein GANs_](https://arxiv.org/abs/1704.00028) and it is assumed that the authors of PassGAN used the [improved_wgan_training](https://github.com/igul222/improved_wgan_training) tensorflow implementation in their work. For this reason, I have modified that reference implementation in this repository to make it easy to train (`train.py`) and sample (`sample.py`) from. This repo contributes:
- A command-line interface
- A pretrained PassGAN model trained on the RockYou dataset
## Getting Started
```bash
# requires CUDA 8 to be pre-installed
pip install -r requirements.txt
```
### Generating password samples
Use the pretrained model to generate 1,000,000 passwords, saving them to `gen_passwords.txt`.
```bash
python sample.py \
--input-dir pretrained \
--checkpoint pretrained/checkpoints/195000.ckpt \
--output gen_passwords.txt \
--batch-size 1024 \
--num-samples 1000000
```
### Training your own models
Training a model on a large dataset (100MB+) can take several hours on a GTX 1080.
```bash
# download the rockyou training data
# contains 80% of the full rockyou passwords (with repeats)
# that are 10 characters or less
curl -L -o data/train.txt https://github.com/brannondorsey/PassGAN/releases/download/data/rockyou-train.txt
# train for 200000 iterations, saving checkpoints every 5000
# uses the default hyperparameters from the paper
python train.py --output-dir output --training-data data/train.txt
```
You are encouraged to train using your own password leaks and datasets. Some great places to find those include:
- [LinkedIn leak](https://github.com/brannondorsey/PassGAN/releases/download/data/68_linkedin_found_hash_plain.txt.zip) (1.7GB compressed, direct download. Mirror from [Hashes.org](https://hashes.org/leaks.php))
- [Exploit.in torrent](https://thepiratebay.org/torrent/16016494/exploit.in) (10GB+, 800 million accounts. Infamous!)
- [Hashes.org](https://hashes.org/leaks.php): Awesome shared password recovery site. Consider donating if you have the resources ;)
## Results
I've yet to do an exhaustive analysis of my attempt to reproduce the results from the PassGAN paper. However, using the pretrained rockyou model to generate 10⁸ password samples I was able to match 630,347 (23.97%) unique passwords in the test data, using a 80%/20% train/test split.
In general, I am somewhat surprised (and dissapointed) that the authors of PassGAN referenced [prior work](https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_melicher.pdf) in the ML password generation domain but did not compare their results to that research. My initial experience with PassGAN leads me to believe that it would significantly underperform both the RNN and Markov-based approaches mentioned in that paper and I hope that it is not for this reason that the authors have chosen not to compare results.
## Attribution and License
This code is released under an [MIT License](https://github.com/igul222/improved_wgan_training/blob/master/LICENSE). You are free to use, modify, distribute, or sell it under those terms.
The majority of the credit for the code in this repository goes to @igul222 for his work on the [improved_wgan_training](https://github.com/igul222/improved_wgan_training). I've simply modularized his code a bit, added a command-line interface, and specialized it for the PassGAN paper.
The PassGAN [research and paper](https://arxiv.org/abs/1709.00440) was published by Briland Hitaj, Paolo Gasti, Giuseppe Ateniese, Fernando Perez-Cruz.