kerb-sts is no longer maintained, and this repository will be removed from GitHub on or after Wednesday, March 4, 2020. Published release artifacts will continue to be available indefinitely via PyPI.
Based on the ADFS-CLI script originally posted by Quint Van Deman.
This script provides a seamless mechanism for federating the AWS CLI. When properly configured, this script allows a user to get a short lived (1 hour) set of credentials for each authorized role.
The script leverages Kerberos and a SAML-compatible IdP to avoid any need for the user to enter an AD domain password, or provide AWS credentials. However, users can also authenticate using NTLM with their username and password or with a Kerberos keytab.
See our security policy for handling of security-related matters.
Kerb-STS looks for configuration in the ~/.kerb-sts/config.json file. This file contains the following fields:
| Field | Required? | Description |
|---|---|---|
| idp_url | Yes | URL where the SAML authentication requests are sent |
| adfs_url | No | deprecated URL where the SAML authentication requests are sent |
| region | Yes | Region for AWS credentials |
| kerb_domain | No | Domain name used for the Kerberos GSS exchange. This is set to the domain name of idp_url by default |
| preferred_auth_type | No | Type of authentication to use when a username and password is submitted. Acceptable values: ntlm (default), kerberos |
Users can generate this file with Kerb-STS:
kerb-sts --configure
This will prompt the user for those values and then serialize the configuration. Users can also manually create the configuration file. A sample for AD FS is demonstrated below:
{
"region": "us-east-1",
"idp_url": "https://sample.domain.com/adfs/ls/IdpInitiatedSignOn.aspx?loginToRp=urn:amazon:webservices"
}
Users can override either of the configured values on the command line.
- Note: Python 2.7.10 is the minimal version supported
- Note: All platforms have been tested with both Python 2.7 and 3.5
- Note: If you are using El Capitan or Sierra, refer to the subsequent OSX section
- sudo easy_install pip
- sudo pip install kerb-sts
- Note: El Capitan forces the version of some modules which directly interfere with kerb-sts. In order to get it to work users need to either use a version of Python that was not included with the OS or need to follow these instructions which leverage virtual environments.
- sudo easy_install pip virtualenv
- virtualenv ~/kerb-sts
- source ~/kerb-sts/bin/activate
- sudo pip install kerb-sts --ignore-installed six
- deactivate
- sudo ln -s ~/kerb-sts/bin/kerb-sts /usr/local/bin/kerb-sts
- You will need to update your version of Python to 2.7.12+; Homebrew is the easiest method.
- You will also need to install/update the XCODE Development Extensions 1a. sudo xcode-select install
- You can then just run sudo pip install kerb-sts
- Install Python
- Ensure python and python/scripts are on the PATH
- Install pywin32 from mhammond/pywin32. Follow the instructions to ensure you get the correct version.
- pip install kerb-sts
- sudo apt-get update
- sudo apt-get install -y krb5-kdc libkrb5-dev python-setuptools python3-pip
- sudo pip install kerb-sts
If the install went smoothly kerb-sts should be on your path. There are a lot of configuration options.
The best way to discover them is to check out the help statement.
kerb-sts --help
The script allows users to specify an AWS IAM role that will be set as the default IAM role in the credentials file.
kerb-sts -r [iam-role-to-assume]
All subsequent AWS CLI commands will use this role by default.
Additionally, all available roles will be added as named profiles to the credentials file. Users can then leverage the default role or use the AWS_DEFAULT_PROFILE environment variable to select a specific role/profile. You can find more information about the credentials file in the AWS Documentation.
If you have a lot of roles available, it may take a while to process all available roles.
In this case, you can use the -o option and only the specified role will be processed.
It will also be treated as the default role (-r option).
By passing in a --daemon flag, the script will continue running and update the credentials file every
half hour. The refresh time can be set with the --refresh argument, but remember
the tokens only last for one hour.
kerb-sts -r iam-role-to-assume --daemon
kerb-sts -u username -p 'password' -d DOMAIN
Based on the value of preferred_auth_type in your config file, this call with authenticate using NTLM, or will create a kerberos ticket for the inputted user and authenticate.
This script allows users to generate Kerberos tokens with Kerberos keytabs. Keytabs are private key files that are signed with the user's name, domain, and password. You can generate a keytab by running:
ktutil -k username.keytab add -p [email protected] -e arcfour-hmac-md5 -V 1
Users can use the keytab to authenticate with Kerberos by running:
kinit -kt username.keytab [email protected]
Keytabs allow users to authenticate without their password. The keytab is signed with the password however, so when a password is updated the keytab must likewise be updated. They can then be used with kerb-sts to generate temporary tokens:
kerb-sts --key username.keytab -u username -d DOMAIN.COM
The default location for the AWS credentials file is ~/.aws/credentials. Users are also able to specify a different location for the credentials generated.
kerb-sts -c ./aws-credentials
If you are having issues authenticating with Kerberos, make sure you can run kinit. This should prompt you for
your password and then login successfully. You can view your current Kerberos tickets with klist. If you want to
ensure Kerberos is working properly you can delete all of your tickets with kdestroy -A and then try to get another
ticket issued by running kinit.
The easiest way to install and distribute kerb-sts is using a wheel. A distribution can be built by running:
python setup.py bdist_wheel
That should output a .whl file in the dist directory which can be installed with pip.
Kerb-STS can also be built into a standalone executable with Python bundled to ease installation.
python setup.py install
pip install pyinstaller
pyinstaller --onfile kerb_sts/__main__.py
This will produce a dist/__main__.exe which can then be renamed/run as a standalone exe.
The recommended way to install locally from source is to use a virtual environment. From the root of the kerb-sts source code directory run:
- pip install virtualenv
- virtualenv venv
- source venv/bin/activate
- python setup.py install
- python kerb_sts/