Credential Digger is a GitHub scanning tool that identifies hardcoded credentials (Passwords, API Keys, Secret Keys, Tokens, personal information, etc), filtering the false positive data through machine learning models.
TLDR; watch the video ⬇️
- Why
- Requirements
- Download and installation
- How to run
- Docker container
- Advanced installation
- How to update the project
- Python library usage
- CLI - Command Line Interface
- Wiki
- Contributing
- How to obtain support
- News
In data protection, one of the most critical threats is represented by hardcoded (or plaintext) credentials in open-source projects. Several tools are already available to detect leaks in open-source platforms, but the diversity of credentials (depending on multiple factors such as the programming language, code development conventions, or developers' personal habits) is a bottleneck for the effectiveness of these tools. Their lack of precision leads to a very high number of pieces of code incorrectly detected as leaked secrets. Data wrongly detected as a leak is called false positive data, and compose the huge majority of the data detected by currently available tools.
The goal of Credential Digger is to reduce the amount of false positive data on the output of the scanning phase by leveraging machine learning models.
The tool supports several scan flavors: public and private repositories on github and gitlab, wiki pages, github organizations, local git repositories, local files and folders. Please refer to the Wiki for the complete documentation.
For the complete description of the approach of Credential Digger, you can read this publication.
@InProceedings {lrnto-icissp21,
author = {S. Lounici and M. Rosa and C. M. Negri and S. Trabelsi and M. Önen},
booktitle = {Proc. of the 8th The International Conference on Information Systems Security and Privacy (ICISSP)},
title = {Optimizing Leak Detection in Open-Source Platforms with Machine Learning Techniques},
month = {February},
day = {11-13},
year = {2021}
}
Credential Digger supports Python >= 3.6 and < 3.10, and works only with Linux and MacOS systems. In case you don't meet these requirements, you may consider running a Docker container (that also includes a user interface).
First, you need to install the regular expression matching library Hyperscan. Be sure to have build-essential
and python3-dev
too.
sudo apt install -y libhyperscan-dev build-essential python3-dev
or (for MacOS):
brew install hyperscan
Then, you can install Credential Digger module using pip
.
pip install credentialdigger
One of the core components of Credential Digger is the regular expression scanner. You can choose the regular expressions rules you want (just follow the template here). We provide a list of patterns in the rules.yml
file, that are included in the UI.
Before the very first scan, you need to add the rules that will be used by the scanner. This step is only needed once.
python -m credentialdigger add_rules --sqlite /path/to/data.db /path/to/rules.yaml
Credential Digger leverages machine learning models to filter false positives, especially in the identification of passwords:
-
Path Model: A lot of fake credentials reside in example files such as documentation, examples or test files, since it is very common for developers to provide test code for their projects. The Path Model analyzes the path of each discovery and classifies it as false positive when needed.
-
Snippet Model: Identify the portion of code used to authenticate with passwords, and distinguish between real and dummy passwords. This model is composed of a pre-processing step (Extractor) and a classification step (Classifier).
To install the models, you first need to export them as environment variables, and them download them:
export path_model=https://github.com/SAP/credential-digger/releases/download/PM-v1.0.1/path_model-1.0.1.tar.gz
export snippet_model=https://github.com/SAP/credential-digger/releases/download/SM-v1.0.0/snippet_model-1.0.0.tar.gz
python -m credentialdigger download path_model
python -m credentialdigger download snippet_model
WARNING: Don't run the download command from the installation folder of credentialdigger in order to avoid errors in linking.
WARNING: We provide the pre-trained models, but we do not guarantee the efficiency of these models. If you want more accurate machine learning models, you can train your own models (just replace the binaries with your own models) or use the fine-tuning option.
After adding the rules, you can scan a repository:
python -m credentialdigger scan https://github.com/user/repo --sqlite /path/to/data.db
Machine learning models are not mandatory, but highly recommended in order to reduce the manual effort of reviewing the result of a scan:
python -m credentialdigger scan https://github.com/user/repo --sqlite /path/to/data.db --models PathModel SnippetModel
To have a ready-to-use instance of Credential Digger, with a user interface, you can build the docker container. This option requires the installation of Docker and Docker Compose.
git clone https://github.com/SAP/credential-digger.git
cd credential-digger
cp .env.sample .env
sudo docker-compose up --build
The UI is available at http://localhost:5000/
Credential Digger is modular, and offers a wide choice of components and adaptations.
After installing the dependencies listed above, you can install Credential Digger as follows.
Configure a virtual environment for Python 3 (optional) and clone the main branch of the project:
virtualenv --system-site-packages -p python3 ./venv
source ./venv/bin/activate
git clone https://github.com/SAP/credential-digger.git
cd credential-digger
Install the requirements from requirements.txt
file and install the library:
pip install -r requirements.txt
python setup.py install
Then, you can add the rules, install the machine learning libraries, and scan a repository as described above.
Another ready-to-use instance of Credential Digger with the UI, but using a dockerized postgres database instead of a local sqlite one:
git clone https://github.com/SAP/credential-digger.git
cd credential-digger
cp .env.sample .env
vim .env # set credentials for postgres
sudo docker-compose -f docker-compose.postgres.yml up --build
WARNING: Differently from the sqlite version, here we need to configure the
.env
file with the credentials for postgres (by modifyingPOSTGRES_USER
,POSTGRES_PASSWORD
andPOSTGRES_DB
).
Most advanced users may also wish to use an external postgres database instead of the dockerized one we provide in our docker-compose.postgres.yml
.
If you are already running Credential Digger and you want to update it to a newer version, you can refer to the wiki for the needed steps.
When installing credentialdigger from pip (or from source), you can instantiate the client and scan a repository.
Instantiate the client proper for the chosen database:
# Using a Sqlite database
from credentialdigger import SqliteClient
c = SqliteClient(path='/path/to/data.db')
# Using a postgres database
from credentialdigger import PgClient
c = PgClient(dbname='my_db_name',
dbuser='my_user',
dbpassword='my_password',
dbhost='localhost_or_ip',
dbport=5432)
Add rules before launching your first scan.
c.add_rules_from_file('/path/to/rules.yml')
new_discoveries = c.scan(repo_url='https://github.com/user/repo',
models=['PathModel', 'SnippetModel'],
debug=True)
WARNING: Make sure you add the rules before your first scan.
WARNING: Make sure you download the models before using them in a scan.
Please refer to the Wiki for further information on the arguments.
Credential Digger offers the possibility to fine-tune the snippet model, by retraining a model on each repository scanned.
If you want to activate this option, set generate_snippet_extractor=True
and enable the SnippetModel
when you scan a repository. You need to install the snippet model before using the fine-tuning option.
new_discoveries = c.scan(repo_url='https://github.com/user/repo',
models=['PathModel', 'SnippetModel'],
generate_snippet_extractor=True,
debug=True)
Credential Digger also offers a simple CLI to scan a repository. The CLI supports both sqlite and postgres databases. In case of postgres, you need either to export the credentials needed to connect to the database as environment variables or to setup a .env
file. In case of sqlite, the path of the db must be passed as argument.
Refer to the Wiki for all the supported commands and their usage.
For further information, please refer to the Wiki
We invite your participation to the project through issues and pull requests. Please refer to the Contributing guidelines for how to contribute.
As a first step, we suggest to read the wiki. In case you don't find the answers you need, you can open an issue or contact the maintainers.