debian-mirror-gitlab/doc/user/application_security/dependency_scanning/analyzers.md

type	stage	group	info
reference, howto	Secure	Composition Analysis	To determine the technical writer assigned to the Stage/Group associated with this page, see https://about.gitlab.com/handbook/engineering/ux/technical-writing/#designated-technical-writers

Dependency Scanning Analyzers (ULTIMATE)

Dependency Scanning relies on underlying third-party tools that are wrapped into what we call "Analyzers". An analyzer is a dedicated project that wraps a particular tool to:

• Expose its detection logic.
• Handle its execution.
• Convert its output to the common format.

This is achieved by implementing the common API.

Dependency Scanning supports the following official analyzers:

• bundler-audit
• gemnasium
• gemnasium-maven
• gemnasium-python
• retire.js

The analyzers are published as Docker images, which Dependency Scanning uses to launch dedicated containers for each analysis.

Dependency Scanning is pre-configured with a set of default images that are maintained by GitLab, but users can also integrate their own custom images.

Official default analyzers

Any custom change to the official analyzers can be achieved by using an environment variable in your .gitlab-ci.yml.

Using a custom Docker mirror

You can switch to a custom Docker registry that provides the official analyzer images under a different prefix. For instance, the following instructs Dependency Scanning to pull my-docker-registry/gl-images/gemnasium instead of registry.gitlab.com/gitlab-org/security-products/analyzers/gemnasium. In .gitlab-ci.yml define:

include:
  template: Dependency-Scanning.gitlab-ci.yml

variables:
  SECURE_ANALYZERS_PREFIX: my-docker-registry/gl-images

This configuration requires that your custom registry provides images for all the official analyzers.

Selecting specific analyzers

You can select the official analyzers you want to run. Here's how to enable bundler-audit and gemnasium while disabling all the other default ones. In .gitlab-ci.yml define:

include:
  template: Dependency-Scanning.gitlab-ci.yml

variables:
  DS_DEFAULT_ANALYZERS: "bundler-audit,gemnasium"

bundler-audit runs first. When merging the reports, Dependency Scanning removes the duplicates and keeps the bundler-audit entries.

Disabling default analyzers

Setting DS_DEFAULT_ANALYZERS to an empty string disables all the official default analyzers. In .gitlab-ci.yml define:

include:
  template: Dependency-Scanning.gitlab-ci.yml

variables:
  DS_DEFAULT_ANALYZERS: ""

That's needed when one totally relies on custom analyzers.

Custom analyzers

You can provide your own analyzers by defining CI jobs in your CI configuration. For consistency, you should suffix your custom Dependency Scanning jobs with -dependency_scanning. Here's how to add a scanning job that's based on the Docker image my-docker-registry/analyzers/nuget and generates a Dependency Scanning report gl-dependency-scanning-report.json when /analyzer run is executed. Define the following in .gitlab-ci.yml:

nuget-dependency_scanning:
  image:
    name: "my-docker-registry/analyzers/nuget"
  script:
    - /analyzer run
  artifacts:
    reports:
      dependency_scanning: gl-dependency-scanning-report.json

The Security Scanner Integration documentation explains how to integrate custom security scanners into GitLab.

Analyzers data

The following table lists the data available for each official analyzer.

Property \ Tool	Gemnasium	bundler-audit	Retire.js
Severity	𐄂	✓	✓
Title	✓	✓	✓
File	✓	⚠	✓
Start line	𐄂	𐄂	𐄂
End line	𐄂	𐄂	𐄂
External ID (e.g., CVE)	✓	✓	⚠
URLs	✓	✓	✓
Internal doc/explanation	✓	𐄂	𐄂
Solution	✓	✓	𐄂
Confidence	𐄂	𐄂	𐄂
Affected item (e.g. class or package)	✓	✓	✓
Source code extract	𐄂	𐄂	𐄂
Internal ID	✓	𐄂	𐄂
Date	✓	𐄂	𐄂
Credits	✓	𐄂	𐄂

• ✓ => we have that data
• ⚠ => we have that data, but it's partially reliable, or we need to extract that data from unstructured content
• 𐄂 => we don't have that data, or it would need to develop specific or inefficient/unreliable logic to obtain it.

The values provided by these tools are heterogeneous, so they are sometimes normalized into common values (e.g., severity, confidence, etc).