debian-mirror-gitlab/doc/user/application_security/index.md
2022-05-07 20:08:51 +05:30

35 KiB

stage group info
Secure Static 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/#assignments

Secure your application (ULTIMATE)

GitLab can check your application for security vulnerabilities including:

  • Unauthorized access.
  • Data leaks.
  • Denial of service attacks.

Statistics and details on vulnerabilities are included in the merge request. Providing actionable information before changes are merged enables you to be proactive.

GitLab also provides high-level statistics of vulnerabilities across projects and groups:

  • The Security Dashboard provides a high-level view of vulnerabilities detected in your projects, pipeline, and groups.
  • The Threat Monitoring page provides runtime security metrics for application environments. With the information provided, you can immediately begin risk analysis and remediation.

For an overview of GitLab application security, see Shifting Security Left.

Security scanning tools

GitLab uses the following tools to scan and report known vulnerabilities found in your project.

Secure scanning tool Description
Container Scanning Scan Docker containers for known vulnerabilities.
Dependency List View your project's dependencies and their known vulnerabilities.
Dependency Scanning Analyze your dependencies for known vulnerabilities.
Dynamic Application Security Testing (DAST) Analyze running web applications for known vulnerabilities.
DAST API Analyze running web APIs for known vulnerabilities.
API fuzzing Find unknown bugs and vulnerabilities in web APIs with fuzzing.
Secret Detection Analyze Git history for leaked secrets.
Security Dashboard View vulnerabilities in all your projects and groups.
Static Application Security Testing (SAST) Analyze source code for known vulnerabilities.
Infrastructure as Code (IaC) Scanning Analyze your IaC configuration files for known vulnerabilities.
Coverage fuzzing Find unknown bugs and vulnerabilities with coverage-guided fuzzing.
Cluster Image Scanning Scan Kubernetes clusters for known vulnerabilities.

Vulnerability scanner maintenance

The following vulnerability scanners and their databases are regularly updated:

Secure scanning tool Vulnerabilities database updates
Container Scanning A job runs on a daily basis to build new images with the latest vulnerability database updates from the upstream scanner. For more details, see Vulnerabilities database update.
Dependency Scanning Relies on bundler-audit (for Ruby gems), retire.js (for npm packages), and gemnasium (the GitLab tool for all libraries). Both bundler-audit and retire.js fetch their vulnerabilities data from GitHub repositories, so vulnerabilities added to ruby-advisory-db and retire.js are immediately available. The tools themselves are updated once per month if there's a new version. The Gemnasium DB is updated on a daily basis using data from NVD, the ruby-advisory-db and the GitHub Security Advisory Database as data sources. See our current measurement of time from CVE being issued to our product being updated.
Dynamic Application Security Testing (DAST) The scanning engine is updated on a periodic basis. See the version of the underlying tool zaproxy. The scanning rules are downloaded at scan runtime.
Static Application Security Testing (SAST) Relies exclusively on the tools GitLab wraps. The underlying analyzers are updated at least once per month if a relevant update is available. The vulnerabilities database is updated by the upstream tools.

In versions of GitLab that use the same major version of the analyzer, you do not have to update GitLab to benefit from the latest vulnerabilities definitions. The security tools are released as Docker images. The vendored job definitions that enable them use major release tags according to semantic versioning. Each new release of the tools overrides these tags. Although in a major analyzer version you automatically get the latest versions of the scanning tools, there are some known issues with this approach.

Security scanning with Auto DevOps

To enable all GitLab Security scanning tools, with default settings, enable Auto DevOps:

While you cannot directly customize Auto DevOps, you can include the Auto DevOps template in your project's .gitlab-ci.yml file.

Security scanning without Auto DevOps

To enable all GitLab security scanning tools, with the option of customizing settings, add the GitLab CI/CD templates to your .gitlab-ci.yml file.

To enable Static Application Security Testing, Dependency Scanning, License Scanning, and Secret Detection, add:

include:
  - template: Security/Dependency-Scanning.gitlab-ci.yml
  - template: Security/License-Scanning.gitlab-ci.yml
  - template: Security/SAST.gitlab-ci.yml
  - template: Security/Secret-Detection.gitlab-ci.yml

To enable Dynamic Application Security Testing (DAST) scanning, add the following to your .gitlab-ci.yml. Replace https://staging.example.com with a staging server's web address:

include:
  - template: Security/DAST.gitlab-ci.yml

variables:
  DAST_WEBSITE: https://staging.example.com

For more details about each of the security scanning tools, see their respective documentation sections.

Override the default registry base address

By default, GitLab security scanners use registry.gitlab.com/security-products as the base address for Docker images. You can override this globally by setting the CI/CD variable SECURE_ANALYZERS_PREFIX to another location. Note that this affects all scanners at once.

Use security scanning tools with merge request pipelines

By default, the application security jobs are configured to run for branch pipelines only. To use them with merge request pipelines, you may need to override the default rules: configuration to add:

rules:
  - if: $CI_PIPELINE_SOURCE == "merge_request_event"

Default behavior of GitLab security scanning tools

Secure jobs in your pipeline

If you add the security scanning jobs as described in Security scanning with Auto DevOps or Security scanning without Auto DevOps to your .gitlab-ci.yml each added security scanning tool behave as described below.

For each compatible analyzer, a job is created in the test, dast or fuzz stage of your pipeline and runs on the next new branch pipeline. Features such as the Security Dashboard, Vulnerability Report, and Dependency List that rely on this scan data only show results from pipelines on the default branch, only if all jobs are finished, including manual ones. One tool might use many analyzers.

Our language and package manager specific jobs attempt to assess which analyzer(s) they should run for your project so that you can do less configuration.

If you want to override this to increase the pipeline speed you may choose which analyzers to exclude if you know they are not applicable (languages or package managers not contained in your project) by following variable customization directions for that specific tool.

Secure job status

Jobs pass if they are able to complete a scan. A pass result does NOT indicate if they did, or did not, identify findings. The only exception is coverage fuzzing, which fails if it identifies findings.

Jobs fail if they are unable to complete a scan. You can view the pipeline logs for more information.

All jobs are permitted to fail by default. This means that if they fail it do not fail the pipeline.

If you want to prevent vulnerabilities from being merged, you should do this by adding Security Approvals in Merge Requests which prevents unknown, high or critical findings from being merged without an approval from a specific group of people that you choose.

We do not recommend changing the job allow_failure setting as that fails the entire pipeline.

JSON Artifact

The artifact generated by the secure analyzer contains all findings it discovers on the target branch, regardless of whether they were previously found, dismissed, or completely new (it puts in everything that it finds).

View security scan information in merge requests (FREE)

All tiers

Merge requests which have run security scans let you know that the generated reports are available to download. To download a report, select Download results, and select the desired report.

Security widget

Ultimate

A merge request contains a security widget which displays a summary of the NEW results. New results are determined by comparing the current findings against existing findings in the target (default) branch (if there are prior findings).

We recommended you run a scan of the default branch before enabling feature branch scans for your developers. Otherwise, there is no base for comparison and all feature branches display the full scan results in the merge request security widget.

The merge request security widget displays only a subset of the vulnerabilities in the generated JSON artifact because it contains both NEW and EXISTING findings.

From the merge request security widget, select Expand to unfold the widget, displaying any new and no longer detected (removed) findings by scan type. Select View full report to go directly to the Security tab in the latest branch pipeline.

Security scanning results in a merge request

View security scan information in the pipeline Security tab

A pipeline's security tab lists all findings in the current branch. It includes new findings introduced by this branch and existing vulnerabilities that were already present when the branch was created. These results likely do not match the findings displayed in the Merge Request security widget as those do not include the existing vulnerabilities (with the exception of showing any existing vulnerabilities that are no longer detected in the feature branch).

For more details, see security tab.

View security scan information in the Security Dashboard

The Security Dashboard show vulnerabilities present in a project's default branch. Data is updated every 24 hours. Vulnerability count updates resulting from any feature branches introducing new vulnerabilities that are merged to default are included after the daily data refresh.

For more details, see Security Dashboard.

View security scan information in the Vulnerability Report

The vulnerability report shows the results of the last completed pipeline on the default branch. It is updated on every pipeline completion. All detected vulnerabilities are shown as well as any previous ones that are no longer detected in the latest scan. Vulnerabilities that are no longer detected may have been remediated or otherwise removed and can be marked as Resolved after proper verification. Vulnerabilities that are no longer detected are denoted with an icon for filtering and review.

By default, the vulnerability report does not show vulnerabilities of dismissed or resolved status so you can focus on open vulnerabilities. You can change the Status filter to see these.

Read more about the Vulnerability report.

Security approvals in merge requests

Introduced in GitLab 12.2.

You can enforce an additional approval for merge requests that would introduce one of the following security issues:

Vulnerability-Check rule

WARNING: This feature is in its end-of-life process. It is deprecated for use in GitLab 14.8, and is planned for removal in GitLab 15.0. Users should migrate to the new Security Approval Policies.

To prevent a merge request introducing a security vulnerability in a project, enable the Vulnerability-Check rule. While this rule is enabled, additional merge request approval by eligible approvers is required when the latest security report in a merge request:

  • Contains vulnerabilities with states (for example, previously detected, dismissed) matching the rule's vulnerability states. Only newly detected are considered if the target branch differs from the project default branch.
  • Contains vulnerabilities with severity levels (for example, high, critical, or unknown) matching the rule's severity levels.
  • Contains a vulnerability count higher than the rule allows.
  • Is not yet generated (until pipeline completion).

An approval is optional when the security report:

  • Contains only vulnerabilities with states (for example, newly detected, resolved) NOT matching the rule's vulnerability states.
  • Contains only vulnerabilities with severity levels (for example, low, medium) NOT matching the rule's severity levels.
  • Contains a vulnerability count equal to or less than what the rule allows.

Project members with at least the Maintainer role can enable or edit the Vulnerability-Check rule.

Enable the Vulnerability-Check rule

To enable or edit the Vulnerability-Check rule:

  1. On the top bar, select Menu > Projects and find your project.
  2. On the left sidebar, select Settings > General.
  3. Expand Merge request approvals.
  4. Select Activate or Edit of the Vulnerability-Check.
  5. Complete the fields. Approvals required must be at least 1.
  6. Select Add approval rule.

The approval rule is enabled for all merge requests. Any code changes reset the approvals required.

Using private Maven repositories

If you have a private Apache Maven repository that requires login credentials, you can use the MAVEN_CLI_OPTS CI/CD variable to pass a username and password. You can set it under your project's settings so that your credentials aren't exposed in .gitlab-ci.yml.

If the username is myuser and the password is verysecret then you would set the following variable under your project's settings:

Type Key Value
Variable MAVEN_CLI_OPTS --settings mysettings.xml -Drepository.password=verysecret -Drepository.user=myuser
<!-- mysettings.xml -->
<settings>
    ...
    <servers>
        <server>
            <id>private_server</id>
            <username>${private.username}</username>
            <password>${private.password}</password>
        </server>
    </servers>
</settings>

Using a custom scanning stage

When security scanning is enabled by including CI/CD templates as described in the Security scanning without Auto DevOps section, the scanning jobs use the predefined test stage by default. If you specify a custom stage in your .gitlab-ci.yml file without including a test stage, an error occurs.

For example, the following attempts to use a unit-tests stage:

include:
  - template: Security/Dependency-Scanning.gitlab-ci.yml
  - template: Security/License-Scanning.gitlab-ci.yml
  - template: Security/SAST.gitlab-ci.yml
  - template: Security/Secret-Detection.gitlab-ci.yml

stages:
  - unit-tests

custom job:
  stage: unit-tests
  script:
    - echo "custom job"

The above .gitlab-ci.yml causes a linting error:

Found errors in your .gitlab-ci.yml:
- dependency_scanning job: chosen stage does not exist; available stages are .pre
- unit-tests
- .post

This error appears because the test stage used by the security scanning jobs isn't declared in the .gitlab-ci.yml file. To fix this issue, you can either:

  • Add a test stage in your .gitlab-ci.yml:

    include:
      - template: Security/Dependency-Scanning.gitlab-ci.yml
      - template: Security/License-Scanning.gitlab-ci.yml
      - template: Security/SAST.gitlab-ci.yml
      - template: Security/Secret-Detection.gitlab-ci.yml
    
    stages:
      - test
      - unit-tests
    
    custom job:
      stage: unit-tests
      script:
        - echo "custom job"
    
  • Override the default stage of each security job. For example, to use a pre-defined stage named unit-tests:

    include:
      - template: Security/Dependency-Scanning.gitlab-ci.yml
      - template: Security/License-Scanning.gitlab-ci.yml
      - template: Security/SAST.gitlab-ci.yml
      - template: Security/Secret-Detection.gitlab-ci.yml
    
    stages:
      - unit-tests
    
    dependency_scanning:
      stage: unit-tests
    
    license_scanning:
      stage: unit-tests
    
    sast:
      stage: unit-tests
    
    .secret-analyzer:
      stage: unit-tests
    
    custom job:
      stage: unit-tests
      script:
        - echo "custom job"
    

Learn more on overriding security jobs:

All the security scanning tools define their stage, so this error can occur with all of them.

Security report validation

  • Introduced in GitLab 13.11.
  • Schema validation message added in GitLab 14.0.

You can enforce validation of the security report artifacts before ingesting the vulnerabilities. This prevents ingestion of broken vulnerability data into the database. GitLab validates the artifacts based on the report schemas.

In GitLab 14.0 and later, when artifact validation is enabled, the pipeline's Security tab lists any report artifacts that failed validation.

Enable security report validation

Deprecated in GitLab 14.9, and planned for removal in GitLab 15.0.

To enable report artifacts validation, set the VALIDATE_SCHEMA environment variable to "true" for the desired jobs in the .gitlab-ci.yml file.

For example, to enable validation for only the sast job:

include:
  - template: Security/Dependency-Scanning.gitlab-ci.yml
  - template: Security/License-Scanning.gitlab-ci.yml
  - template: Security/SAST.gitlab-ci.yml
  - template: Security/Secret-Detection.gitlab-ci.yml
stages:
  - security-scan
dependency_scanning:
  stage: security-scan
license_scanning:
  stage: security-scan
sast:
  stage: security-scan
  variables:
    VALIDATE_SCHEMA: "true"
.secret-analyzer:
  stage: security-scan

Interact with findings and vulnerabilities

You can interact with the results of the security scanning tools in several locations:

For more details about which findings or vulnerabilities you can view in each of those locations, select the respective link. Each page details the ways in which you can interact with the findings and vulnerabilities. As an example, in most cases findings start out as detected status.

You have the option to:

Security scanning configuration tips

Each GitLab security scanning tool has a default CI/CD configuration file, also known as a template.

When customizing the configuration:

  • Include the scanning tool's CI/CD template. Don't copy the content of the template.
  • Use the stable version of each template for production workflows. The stable version changes less often, and breaking changes are only made between major GitLab versions. The latest version contains the most recent changes, but may have significant changes between minor GitLab versions.
  • Only override values in the template as needed. All other values are inherited from the template.

Enforce scan execution

Security and compliance teams must ensure that security scans:

  • Run on a regular basis for all projects.
  • Can't be disabled by developers.

GitLab provides two methods of accomplishing this, each with advantages and disadvantages.

  • Compliance framework pipelines are recommended when:

    • Scan execution enforcement is required for SAST IaC, Dependency Scanning, License Compliance, API Fuzzing, or Coverage-guided Fuzzing.
    • Scan execution enforcement is required for scanners external to GitLab.
    • Enforced execution is required for custom jobs other than security scans.
  • Scan execution policies are recommended when:

    • Scan execution enforcement is required for DAST, SAST, Secret Detection, or Container Scanning.
    • Scans are required to run on a regular, scheduled cadence.

Additional details about the differences between the two solutions are outlined below:

Compliance Framework Pipelines Scan Execution Policies
Flexibility Supports anything that can be done in a CI file. Limited to only the items for which GitLab has explicitly added support. DAST, SAST, Secret Detection, and Container Scanning scans are supported.
Usability Requires knowledge of CI YAML. Follows a rules and actions-based YAML structure.
Inclusion in CI pipeline The compliance pipeline is executed instead of the project's gitlab-ci.yml file. To include the project's gitlab-ci.yml file, use an include statement. Defined variables aren't allowed to be overwritten by the included project's YAML file. Forced inclusion of a new job into the CI pipeline. DAST jobs that must be customized on a per-project basis can have project-level Site Profiles and Scan Profiles defined. To ensure separation of duties, these profiles are immutable when referenced in a scan execution policy. All jobs can be customized as part of the security policy itself with the same variables that are normally available to the CI job.
Schedulable Can be scheduled through a scheduled pipeline on the group. Can be scheduled natively through the policy configuration itself.
Separation of Duties Only group owners can create compliance framework labels. Only project owners can apply compliance framework labels to projects. The ability to make or approve changes to the compliance pipeline definition is limited to individuals who are explicitly given access to the project that contains the compliance pipeline. Only project owners can define a linked security policy project. The ability to make or approve changes to security policies is limited to individuals who are explicitly given access to the security policy project.
Ability to apply one standard to multiple projects The same compliance framework label can be applied to multiple projects inside a group. The same security policy project can be used for multiple projects across GitLab with no requirement of being located in the same group.

Feedback is welcome on our vision for unifying the user experience for these two features

Troubleshooting

Secure job failing with exit code 1

If a Secure job is failing and it's unclear why, add SECURE_LOG_LEVEL: "debug" as a global CI/CD variable for more verbose output that is helpful for troubleshooting.

variables:
  SECURE_LOG_LEVEL: "debug"

Outdated security reports

When a security report generated for a merge request becomes outdated, the merge request shows a warning message in the security widget and prompts you to take an appropriate action.

This can happen in two scenarios:

Source branch is behind the target branch

A security report can be out of date when the most recent common ancestor commit between the target branch and the source branch is not the most recent commit on the target branch.

To fix this issue, rebase or merge to incorporate the changes from the target branch.

Incorporate target branch changes

Target branch security report is out of date

This can happen for many reasons, including failed jobs or new advisories. When the merge request shows that a security report is out of date, you must run a new pipeline on the target branch. Select new pipeline to run a new pipeline.

Run a new pipeline

Getting warning messages … report.json: no matching files

This message is often followed by the error No files to upload, and preceded by other errors or warnings that indicate why the JSON report wasn't generated. Check the entire job log for such messages. If you don't find these messages, retry the failed job after setting SECURE_LOG_LEVEL: "debug" as a custom CI/CD variable. This provides extra information to investigate further.

Getting error message sast job: config key may not be used with 'rules': only/except

When including a .gitlab-ci.yml template like SAST.gitlab-ci.yml, the following error may occur, depending on your GitLab CI/CD configuration:

Found errors in your .gitlab-ci.yml:

    jobs:sast config key may not be used with `rules`: only/except

This error appears when the included job's rules configuration has been overridden with the deprecated only or except syntax. To fix this issue, you must either:

Learn more on overriding SAST jobs.

Transitioning your only/except syntax to rules

When overriding the template to control job execution, previous instances of only or except are no longer compatible and must be transitioned to the rules syntax.

If your override is aimed at limiting jobs to only run on master, the previous syntax would look similar to:

include:
  - template: Security/SAST.gitlab-ci.yml

# Ensure that the scanning is only executed on master or merge requests
spotbugs-sast:
  only:
    refs:
      - master
      - merge_requests

To transition the above configuration to the new rules syntax, the override would be written as follows:

include:
  - template: Security/SAST.gitlab-ci.yml

# Ensure that the scanning is only executed on master or merge requests
spotbugs-sast:
  rules:
    - if: $CI_COMMIT_BRANCH == "master"
    - if: $CI_MERGE_REQUEST_ID

If your override is aimed at limiting jobs to only run on branches, not tags, it would look similar to:

include:
  - template: Security/SAST.gitlab-ci.yml

# Ensure that the scanning is not executed on tags
spotbugs-sast:
  except:
    - tags

To transition to the new rules syntax, the override would be rewritten as:

include:
  - template: Security/SAST.gitlab-ci.yml

# Ensure that the scanning is not executed on tags
spotbugs-sast:
  rules:
    - if: $CI_COMMIT_TAG == null

Learn more on the usage of rules.

Pin your templates to the deprecated versions

To ensure the latest support, we strongly recommend that you migrate to rules.

If you're unable to immediately update your CI configuration, there are several workarounds that involve pinning to the previous template versions, for example:

include:
  remote: 'https://gitlab.com/gitlab-org/gitlab/-/raw/12-10-stable-ee/lib/gitlab/ci/templates/Security/SAST.gitlab-ci.yml'

Additionally, we provide a dedicated project containing the versioned legacy templates. This can be used for offline setups or anyone wishing to use Auto DevOps.

Instructions are available in the legacy template project.

Vulnerabilities are found, but the job succeeds. How can I have a pipeline fail instead?

In these circumstances, that the job succeeds is the default behavior. The job's status indicates success or failure of the analyzer itself. Analyzer results are displayed in the job logs, Merge Request widget or Security Dashboard.

Error: job is used for configuration only, and its script should not be executed

Changes made in GitLab 13.4 to the Security/Dependency-Scanning.gitlab-ci.yml and Security/SAST.gitlab-ci.yml templates mean that if you enable the sast or dependency_scanning jobs by setting the rules attribute, they fail with the error (job) is used for configuration only, and its script should not be executed.

The sast or dependency_scanning stanzas can be used to make changes to all SAST or Dependency Scanning, such as changing variables or the stage, but they cannot be used to define shared rules.

There is an issue open to improve extendability. Please upvote the issue to help with prioritization, and contributions are welcomed.

Empty Vulnerability Report, Dependency List, License list pages

If the pipeline has manual steps with a job that has the allow_failure: false option, and this job is not finished, GitLab can't populate listed pages with the data from security reports. In this case, the Vulnerability Report, the Dependency List, and the License list pages will be empty. These security pages can be populated by running the jobs from the manual step of the pipeline.

There is an issue open to handle this scenario. Please upvote the issue to help with prioritization, and contributions are welcomed.