/gcp_scanner

Primary LanguagePythonApache License 2.0Apache-2.0

pytests

Disclaimer

This project is not an official Google project. It is not supported by Google and Google specifically disclaims all warranties as to its quality, merchantability, or fitness for a particular purpose.

GCP Scanner

This is a GCP resource scanner that can help determine what level of access certain credentials possess on GCP. The scanner is designed to help security engineers with evaluating the impact of a certain VM/container compromise, GCP service account or OAuth2 token key leak.

Currently, the scanner supports the following GCP resources:

  • GCE
  • GCS
  • GKE
  • AppEngine
  • CloudSQL
  • BigQuery
  • Spanner
  • PubSub
  • CloudFunctions
  • BigTable
  • CloudStore
  • KMS
  • Cloud Services
  • The scanner supports SA impersonation

The scanner supports extracting and using the following types of credentials:

  • GCP VM instance metadata;
  • User credentials stored in gcloud profiles;
  • OAuth2 Refresh Token with cloud-platform scope granted;
  • GCP service account key in JSON format.

The scanner does not rely on any third-party tool (e.g. gcloud). Thus, it can be compiled as a standalone tool and be used on a machine with no GCP SDK installed (e.g. Kubernetes pod). However, please keep in mind that the only OS that is currently supported is Linux.

Please note that GCP offers Policy Analyzer to find out which principals (users, service accounts, groups, and domains), have what access to which Google Cloud resources. However, it requires specific permissions on the GCP project and Cloud Assets API needs to be enabled. If you just have a GCP SA key, access to a previously compromised VM, or an OAUth2 refresh token, gcp_scanner is the best option to use.

Installation

To install the package, use pip (you must also have git installed):

pip install https://github.com/google/gcp_scanner/archive/main.zip
gcp-scanner --help

Alternatively:

git clone https://github.com/google/gcp_scanner
cd gcp_scanner
pip install .
gcp-scanner --help

There is a docker build file if you want to run the scanner from a container: docker build -f Dockerfile -t sa_scanner .

Command-line options

usage: gcp-scanner -o /folder_to_save_results/ -g -

optional arguments:
  -h, --help            show this help message and exit
  -k KEY_PATH, --sa_key_path KEY_PATH
                        Path to directory with SA keys in json format
  -g GCLOUD_PROFILE_PATH, --gcloud_profile_path GCLOUD_PROFILE_PATH
                        Path to directory with gcloud profile. Specify - to search in default gcloud config path
  -m                    Extract credentials from GCE instance metadata
  -at ACCESS_TOKEN      Use access token directly to scan GCP resources. Limited by TTL
  -rt REFRESH_TOKEN_FILES
                        A list of comma separated files with refresh_token, client_id, token_uri and client_secret
  -s KEY_NAME           Name of individual SA to scan
  -p TARGET_PROJECT     Name of individual project to scan
  -f FORCE_PROJECTS     Comma separated list of project names to include in the scan
  -c SCAN_CONFIG        A path to configuration file with a set of specific resources to scan.
  -l LOG_LEVEL, -logging LOG_LEVEL
                        Set logging level (INFO, WARNING, ERROR)

Required parameters:
  -o OUTPUT             Path to output directory

Option -f requires an additional explanation. In some cases, the service account does not have permissions to explicitly list project names. However, it still might have access to underlying resources if we provide correct project name. This option is specifically designed to handle such cases.

Building standalone binary with pyinstaller

Please replace google-api-python-client==2.9.0 with google-api-python-client==1.8.0 in pyproject.toml. After that, navigate to the scanner source code directory and use pyinstaller to compile a standalone binary:

pyinstaller -F --add-data 'roots.pem:grpc/_cython/_credentials/" scanner.py

Working with results

The GCP Scanner produces a standard JSON file that can be handled by any JSON Viewer or DB. If you just need a convenient way to grep JSON results, we can recommend gron.

Contributing

See CONTRIBUTING.md for details.

License

Apache 2.0; see LICENSE for details.