Skip to content

A benchmark to compare (compliance) scanners for Kubernetes workloads (i.e. Misconfigurations of workloads and NOT infrastructure security) The tool generates a suite of manifests using cdk8s, where every manifest is considered a single check in the benchmark. The design idea is to have a hardened configurations as the base and then for each che…

License

Notifications You must be signed in to change notification settings

dynatrace-oss/Kalm-Benchmark

πŸ“Š Kalm Benchmark

The goal of the Kalm benchmark is to provide a proper baseline for comparing Kubernetes (workload) compliance/misconfiguration scanners.
This benchmark consists of two parts:

  • a set of manifests, which contain misconfigurations, which scanners should detect
  • a web UI to help with the comparison and analysis of the scanner results on this benchmark
⚠️ This product is not officially supported by Dynatrace.

Description

Benchmark Manifests

To facilitate a fair comparison, a set of manifests is generated. By design the generated manifests follow all the best practices. A single manifest is supposed to trigger a single check of the scanner. The ID and further information for the check are provided as labels and annotations on the generated resources.

This way, it is easy to compare the expected results with the actual results of a scanner. An overview of all the checks in the benchmark can be found in Benchmark Checks.

Web UI

The web application consists of two pages:

  • an overview of various scanners checked with this benchmark
  • an analysis page to inspect the results of a specific scanner in more detail

Use-Cases

  • for Security and Ops teams:

    • find a suitable scanner for your needs
    • aids you in the development of your custom checks/policies/queries/rules for a tool
  • For scanner developers

    • compare your tool to alternatives on the market
    • develop and maintain your implemented checks, e.g.:
      • identify noisy checks (e.g., multiple checks which raise an alert for the same issue)
      • or faulty checks (e.g., alerts which were erroneously raised or not raised due to a bug)

Prerequisites

  • Python >= 3.9
  • The manifests are generated using cdk8s, which depends on nodeJS
    • Please ensure nodeJS is installed on your system
  • Any scanner for which a scan should be triggered must be installed manually
  • Poetry is used to manage the project itself

Getting Started

1) πŸ”¨ Installation

To install the benchmark along with its dependencies listed in the pyproject.toml execute:

poetry install

2) πŸ„β€β™€οΈ Usage

To use this package run the CLI with the appropriate command:

poetry run cli <command>

For detailed information of the available commands or general help run:

poetry run cli --help

2.1) Generating manifests

To generate manifests use the generate command:

poetry run cli generate [--out <target directory>]

These manifests form the basis for the benchmark and will be placed in the directory specified with the --out argument. The location defaults to the manifests folder in the working directory.

2.2) Start the Web UI

Besides the CLI commands the tool also provides a web user interface to manage the scan(s) and analyse evaluation results. It can be started with the command:

poetry run cli serve

2.3) Perform a scan with a Scanner

To scan either a cluster or manifest files with the specified tool use the scan command. Use either the -c flag to specify the target cluster or the -f flag to specify the target file/folder.

poetry run cli scan <tool> [-c | -f <target file or folder>]

❗️ Important executing a scan requires the respective tool to be installed on the system!

E.g., to scan manifests with the tool dummy located in the manifests folder execute:

poetry run cli scan dummy -f manifests

In order to save the results, add the -o flag with the name of the destination folder:

poetry run cli scan <tool> [-c | -f <target file or folder>] -o <output-folder>

2.4) Evaluate a Scanner

For the evaluation of a scanner the tool must be run beforehand and the results must be stored as <tool>_v<version>_<yyyy-mm-dd>.json. Then the stored results can be loaded by the tool for the evaulation.

poetry run cli evaluate <tool>

πŸš€ Deployment

Some scanners only scan resources deployed in a Kubernetes cluster. You can find instructions on how to deploy the benchmark in a cluster here

Tool-specific considerations

kube-bench

As their description states, it focuses on infrastructure security and not workload security:

Checks whether Kubernetes is deployed according to security best practices as defined in the CIS Kubernetes Benchmark.

Thus, the tool is listed in the benchmark just for completness sake and is expected to perform poorly because it focuses on another domain, than what is covered by this benchmark!

KubiScan

KubiScan is not formally installed on a system. Instead, the project can be cloned and executed as Python script. For a consistent interface, the benchmark assumes to that KubiScan can be called directly using the alias kubiscan E.g. to use a dedicated virtual environment (called venv inside the kubiscan directory) you can create the alias in Linux using this command

alias kubiscan="<PATH-TO-DIR>/venv/bin/python <PATH-TO-DIR>/KubiScan.py"

Troubleshooting

The results from docker run can't be parsed

  • ensure the -t flag is not used in the command. If it is, stdout and stderr are joined to just stdout. This means errors can't be handled properly and it corrupts the results in stdout.

πŸ’ͺ Contributing

Want to contribute? Awesome! We welcome contributions of all kinds: new scanners, fixes to the existing implementations, bug reports, and feedback of any kind.


License

Apache Version 2.0

About

A benchmark to compare (compliance) scanners for Kubernetes workloads (i.e. Misconfigurations of workloads and NOT infrastructure security) The tool generates a suite of manifests using cdk8s, where every manifest is considered a single check in the benchmark. The design idea is to have a hardened configurations as the base and then for each che…

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •