Home

Awesome

Password Guessing Framework

The Password Guessing Framework is an open source tool to provide an automated and reliable way to compare password guessers. It can help to identify individual strengths and weaknesses of a guesser, its modes of operation or even the underlying guessing strategies. Therefor, it gathers information about how many passwords from an input file (password leak) have been cracked in relation to the amount of generated guesses. Subsequent to the guessing process an analysis of the cracked passwords is performed.

By default the framework supports the following password guessers:

In general though, any guesser that prints the password candidates via STDOUT can be used with the framework.
Please note: The aforementioned password guessing / password cracking software is not part nor shipped with the framework and need to be installed separately.

Changelog

Refer to docs/CHANGELOG.md for more information.
This software is under active development.
More information on this topic can be found on the project website (www.password-guessing.org).

Dependencies

The framework is written in Python 2.7 and thus requires it installed on the host system.
In addition, the following modules have to be installed:

We have tested the framework on Ubuntu 14.04 where all the requirements are met by default.

Installation

On the project website (www.password-guessing.org) you can find a more extensive installation tutorial, a beginners' guide, use case descriptions, guesser installation tutorials, dataset descriptions, experiments, and measurement data, very soon!

We have created the directory /opt/pgf in which we installed an instance of all guessers we run with the framework (e.g., /opt/pgf/prince).
This way, our individual configurations of the guessers wouldn't collide with other guessing instances installed on the system!
Important to note is that John the Ripper is used for the hash evaluation in cases where hashed input is provided.
If you consider to run John the Ripper Markov mode as password guesser, you will need two instllations on your hard disk.
Therefor, install it in /opt/pgf/john-hash and /opt/pgf/john-guess, respectively.
The /opt/pgf/john-hash version, which is utilized as hash evaluator by the framework requires some additional care.

IMPORTANT NOTE: You have to change two parameters in the john.conf file in your /opt/pgf/john-hash directory:

Configuration / Execution

The framework processes a queue of preconfigured "jobs" which are defined in the configuration file run.ini. A job basically means the execution of a certain password guesser. The guessers are preconfigured in the shell scripts in the folder scripts. The scripts execute the training command (if provided by the guesser) and the guessing command. To configure a job, you find detailed descriptions about the different parameters in the comments at the top of the run.ini. Some parameters are required for all jobs, others are optional. The default values for the optional parameters can be found in the DEFAULT section of the file. Furthermore, there are a couple of special parameters which are used for features independent from any job. These parameters are also to be set in the DEFAULT section.

Once set up, a job execution is split into 3 phases:

In addition to the regular framework features, we provide a web frontend which uses live data to plot guess-number graphs and shows the progress of the current job and the overall progress of the PGF run.

Before you run the framework we suggest to test the single "Guesser Wrapper Scripts", e.g., via ./scripts/PRINCE.sh /opt/pgf/leaks/myspace_training.txt 10.
If you can see the generated passwords, you should give it a try to execute the framework simply by running python main.py.
If you have trouble with the "Guesser Wrapper Scripts" make sure all directories are correctly setup and check whether the scripts are executable (chmod +x /scripts/PRINCE.sh).

Extension

The framework is designed to be easily extendable for specific needs and use cases.
Mainly the following components can be extended:

By default, the Password Guessing Framework supports the input formats pure plaintext and pure hash, meaning input files with one password/hash value per line. Also the widely spread withcount leaks can be used as input files for the framework without any adaptations. Parsers for other input formats, e.g., shadow hash files, can easily be implemented.

For the extension of analysis capabilities, the analysis module provides a method to execute plugin-code which is run after all candidates have been processed.

Alt text

Output

The framework creates a couple of output files for a run in the local results folder. Besides the log.txt which includes log and debug messages, a jobs.json is created which holds information about the configured jobs of the current run. It will be overwritten for each new PGF run but when backing up the output files after a successful run, the jobs.json receives a timestamp and an unique identifier to be able to connect it with the rest of the output files of that run.

The progress and the cracking success is written into a CSV-file for each guesser, respectively. The files serve on the one hand as input for the live visualization module, on the other hand they open the ability to plot detailed Guess-number graphs with GnuPlot or similar software after the comparision is done. The actualization interval is set to 1000 by default, meaning every 1000 candidates the file is updated with the amount of processed candidates, the amount of cracked passwords and the percentage in regard of the total passwords in the input file (leak).

Web Frontend

Web Frontend

To use the optional web visualization frontend, simply follow these steps:

The frontend shows the progress of the current PGF run along with information about the cracking success, the job queue etc.

License

The Password Guessing Framework is licensed under the MIT license.
Please note: The dynamic visualization module utilizes the Highcharts JS (by Torstein Honsi 2014) library which is only free of use in non-commercial software projects. Refer to docs/LICENSE for more information.