Dependency Scanning (ULTIMATE)

The Dependency Scanning feature can automatically find security vulnerabilities in your software dependencies while you're developing and testing your applications. For example, dependency scanning lets you know if your application uses an external (open source) library that is known to be vulnerable. You can then take action to protect your application.

Dependency Scanning is often considered part of Software Composition Analysis (SCA). SCA can contain aspects of inspecting the items your code uses. These items typically include application and system dependencies that are almost always imported from external sources, rather than sourced from items you wrote yourself.

If you're using GitLab CI/CD, you can use dependency scanning to analyze your dependencies for known vulnerabilities. GitLab scans all dependencies, including transitive dependencies (also known as nested dependencies). You can take advantage of dependency scanning by either:

• Including the dependency scanning template in your existing .gitlab-ci.yml file.
• Implicitly using the auto dependency scanning provided by Auto DevOps.

GitLab checks the dependency scanning report, compares the found vulnerabilities between the source and target branches, and shows the information on the merge request. The results are sorted by the severity of the vulnerability.

For an overview, see Dependency Scanning.

Dependency Scanning compared to Container Scanning

GitLab offers both Dependency Scanning and Container Scanning to ensure coverage for all of these dependency types. To cover as much of your risk area as possible, we encourage you to use all of our security scanning tools:

• Dependency Scanning analyzes your project and tells you which software dependencies, including upstream dependencies, have been included in your project, and what known risks the dependencies contain. Dependency Scanning modifies its behavior based on the language and package manager of the project. It typically looks for a lock file then performs a build to fetch upstream dependency information. In the case of containers, Dependency Scanning uses the compatible manifest and reports only these declared software dependencies (and those installed as a sub-dependency). Dependency Scanning cannot detect software dependencies that are pre-bundled into the container's base image. To identify pre-bundled dependencies, enable Container Scanning language scanning using the CS_DISABLE_LANGUAGE_VULNERABILITY_SCAN variable.
• Container Scanning analyzes your containers and tells you about known risks in the operating system's (OS) packages. You can configure it to also report on software and language dependencies, if you enable it and use the CS_DISABLE_LANGUAGE_VULNERABILITY_SCAN variable. Turning this variable on can result in some duplicate findings, as we do not yet de-duplicate results between Container Scanning and Dependency Scanning. For more details, efforts to de-duplicate these findings can be tracked in this epic.

The following table summarizes which types of dependencies each scanning tool can detect:

1. Lock file must be present in the image to be detected.
2. Binary file must be present in the image to be detected.
3. Only when using Trivy

Requirements

Dependency Scanning runs in the test stage, which is available by default. If you redefine the stages in the .gitlab-ci.yml file, the test stage is required.

To run dependency scanning jobs, by default, you need GitLab Runner with the docker or kubernetes executor. If you're using the shared runners on GitLab.com, this is enabled by default. The analyzer images provided are for the Linux/amd64 architecture.

WARNING: If you use your own runners, make sure your installed version of Docker is not 19.03.0. See troubleshooting information for details.

WARNING: Dependency Scanning does not support run-time installation of compilers and interpreters. If you need it, explain why by filling out the survey.

Supported languages and package managers

Dependency Scanning automatically detects the languages used in the repository. All analyzers matching the detected languages are run. There is usually no need to customize the selection of analyzers. We recommend not specifying the analyzers so you automatically use the full selection for best coverage, avoiding the need to make adjustments when there are deprecations or removals. However, you can override the selection using the variable DS_EXCLUDED_ANALYZERS.

The language detection relies on CI job rules and searches a maximum of two directory levels from the repository's root. For example, the gemnasium-dependency_scanning job is enabled if a repository contains either Gemfile, api/Gemfile, or api/client/Gemfile, but not if the only supported dependency file is api/v1/client/Gemfile.

For Java and Python, when a supported dependency file is detected, Dependency Scanning attempts to build the project and execute some Java or Python commands to get the list of dependencies. For all other projects, the lock file is parsed to obtain the list of dependencies without needing to build the project first.

When a supported dependency file is detected, all dependencies, including transitive dependencies are analyzed. There is no limit to the depth of nested or transitive dependencies that are analyzed.

The following languages and dependency managers are supported:

1. Support for Kotlin projects for Android is tracked in issue 336866.

2. Support for these versions of Java is deprecated and is planned to be removed in the GitLab 16.0 release. Additionally, these versions of Java are not supported by the FIPS-enabled image of gemnasium-maven. Official support is limited to LTS versions only. Although it may be possible to use Dependency Scanning with other versions by building a custom dependency scanning image, this approach is not officially supported by GitLab.

3. Gradle is not supported when FIPS mode is enabled.

4. Support for pnpm lockfiles was introduced in GitLab 15.11. pnpm lockfiles do not store bundled dependencies, so the reported dependencies may differ from npm or yarn.

5. The presence of a Pipfile.lock file alone will not trigger the analyzer; the presence of a Pipfile is still required in order for the analyzer to be executed. However, if a Pipfile.lock file is found, it is used by Gemnasium to scan the exact package versions listed in this file.

   Support for Pipfile.lock files without requiring the presence of a Pipfile is tracked in issue: Dependency Scanning of Pipfile.lock without installing project dependencies.

6. Support for Poetry projects with a poetry.lock file was added in GitLab 15.0. Support for projects without a poetry.lock file is tracked in issue: Poetry's pyproject.toml support for dependency scanning.

7. Support for sbt 1.3 and above was added in GitLab 13.9.

How analyzers obtain dependency information

GitLab analyzers obtain dependency information using one of the following two methods:

1. Parsing lockfiles directly.
2. Running a package manager or build tool to generate a dependency information file which is then parsed.

Obtaining dependency information by parsing lockfiles

The following package managers use lockfiles that GitLab analyzers are capable of parsing directly:

1. Dependency Scanning only parses go.sum if it's unable to generate the build list used by the Go project.

2. Support for lockfileVersion = 3 was introduced in GitLab 15.7.

3. Support for Yarn v2 and v3 was introduced in GitLab 15.11. However, this feature is also available to versions of GitLab 15.0 and later.

   The following features are not supported for Yarn v2 or v3:

   • workspaces
   • yarn patch

   Yarn files that contain a patch, a workspace, or both, are still processed, but these features are ignored.

Obtaining dependency information by running a package manager to generate a parsable file

To support the following package managers, the GitLab analyzers proceed in two steps:

1. Execute the package manager or a specific task, to export the dependency information.
2. Parse the exported dependency information.

1. This test uses the default version of maven specified by the `.tool-versions` file.

2. Different versions of Java require different versions of Gradle. The versions of Gradle listed in the above table are pre-installed in the analyzer image. The version of Gradle used by the analyzer depends on whether your project uses a gradlew (Gradle wrapper) file or not:

   • If your project does not use a gradlew file, then the analyzer automatically switches to one of the pre-installed Gradle versions, based on the version of Java specified by the DS_JAVA_VERSION variable. By default, the analyzer uses Java 17 and Gradle 7.3.3.

     For Java versions 8 and 11, Gradle 6.7.1 is automatically selected, and for Java versions 13 to 17, Gradle 7.3.3 is automatically selected.

   • If your project does use a gradlew file, then the version of Gradle pre-installed in the analyzer image is ignored, and the version specified in your gradlew file is used instead.

3. This test confirms that if a Pipfile.lock file is found, it is used by Gemnasium to scan the exact package versions listed in this file.

4. Because of the implementation of go build, the Go build process requires network access, a pre-loaded mod cache via go mod download, or vendored dependencies. For more information, refer to the Go documentation on compiling packages and dependencies.

How analyzers are triggered

GitLab relies on rules:exists to start the relevant analyzers for the languages detected by the presence of the Supported files in the repository as shown in the table above.

The current detection logic limits the maximum search depth to two levels. For example, the gemnasium-dependency_scanning job is enabled if a repository contains either a Gemfile.lock, api/Gemfile.lock, or api/client/Gemfile.lock, but not if the only supported dependency file is api/v1/client/Gemfile.lock.

When a supported dependency file is detected, all dependencies, including transitive dependencies are analyzed. There is no limit to the depth of nested or transitive dependencies that are analyzed.

How multiple files are processed

NOTE: If you've run into problems while scanning multiple files, contribute a comment to this issue.

Python

We only execute one installation in the directory where either a requirements file or a lock file has been detected. Dependencies are only analyzed by gemnasium-python for the first file that is detected. Files are searched for in the following order:

1. requirements.txt, requirements.pip, or requires.txt for projects using Pip.
2. Pipfile or Pipfile.lock for projects using Pipenv.
3. poetry.lock for projects using Poetry.
4. setup.py for project using Setuptools.

The search begins with the root directory and then continues with subdirectories if no builds are found in the root directory. Consequently a Poetry lock file in the root directory would be detected before a Pipenv file in a subdirectory.

Java and Scala

We only execute one build in the directory where a build file has been detected. For large projects that include multiple Gradle, Maven, or sbt builds, or any combination of these, gemnasium-maven only analyzes dependencies for the first build file that is detected. Build files are searched for in the following order:

1. pom.xml for single or multi-module Maven projects.
2. build.gradle or build.gradle.kts for single or multi-project Gradle builds.
3. build.sbt for single or multi-project sbt builds.

The search begins with the root directory and then continues with subdirectories if no builds are found in the root directory. Consequently an sbt build file in the root directory would be detected before a Gradle build file in a subdirectory.

JavaScript

The following analyzers are executed, each of which have different behavior when processing multiple files:

• Gemnasium

  Supports multiple lockfiles

• Retire.js

  Does not support multiple lockfiles. When multiple lockfiles exist, Retire.js analyzes the first lockfile discovered while traversing the directory tree in alphabetical order.

From GitLab 14.8 the gemnasium analyzer scans supported JavaScript projects for vendored libraries (that is, those checked into the project but not managed by the package manager).

Go

Multiple files are supported. When a go.mod file is detected, the analyzer attempts to generate a build list using Minimal Version Selection. If a non-fatal error is encountered, the analyzer falls back to parsing the available go.sum file. The process is repeated for every detected go.mod and go.sum file.

PHP, C, C++, .NET, C#, Ruby, JavaScript

The analyzer for these languages supports multiple lockfiles.

Support for additional languages

Support for additional languages, dependency managers, and dependency files are tracked in the following issues:

Contribute your scanner

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

Configuration

To enable dependency scanning for GitLab 11.9 and later, you must include the Dependency-Scanning.gitlab-ci.yml template that is provided as a part of your GitLab installation. For GitLab versions earlier than 11.9, you can copy and use the job as defined that template.

Add the following to your .gitlab-ci.yml file:

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

The included template creates dependency scanning jobs in your CI/CD pipeline and scans your project's source code for possible vulnerabilities. The results are saved as a dependency scanning report artifact that you can later download and analyze. Due to implementation limitations, we always take the latest dependency scanning artifact available.

Enable Dependency Scanning via an automatic merge request

• Introduced in GitLab 14.1 with a flag named sec_dependency_scanning_ui_enable. Enabled by default.
• Enabled on self-managed in GitLab 14.1.
• Feature flag sec_dependency_scanning_ui_enable removed in GitLab 14.2.

To enable Dependency Scanning in a project, you can create a merge request:

1. On the top bar, select Main menu > Projects and find your project.
2. On the left sidebar, select Security and Compliance > Security configuration.
3. In the Dependency Scanning row, select Configure with a merge request.
4. Review and merge the merge request to enable Dependency Scanning.

Pipelines now include a dependency scanning job.

Customizing the dependency scanning settings

The Dependency Scanning settings can be changed through CI/CD variables by using the variables parameter in .gitlab-ci.yml. For example:

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

variables:
  SECURE_LOG_LEVEL: error

Because template is evaluated before the pipeline configuration, the last mention of the variable takes precedence.

Overriding dependency scanning jobs

WARNING: Beginning in GitLab 13.0, the use of only and except is no longer supported. When overriding the template, you must use rules instead.

To override a job definition (for example, to change properties like variables or dependencies), declare a new job with the same name as the one to override. Place this new job after the template inclusion and specify any additional keys under it. For example, this disables DS_REMEDIATE for the gemnasium analyzer:

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

gemnasium-dependency_scanning:
  variables:
    DS_REMEDIATE: "false"

To override the dependencies: [] attribute, add an override job as above, targeting this attribute:

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

gemnasium-dependency_scanning:
  dependencies: ["build"]

Available CI/CD variables

Dependency scanning can be configured using environment variables.

WARNING: All customization of GitLab security scanning tools should be tested in a merge request before merging these changes to the default branch. Failure to do so can give unexpected results, including a large number of false positives.

Configuring dependency scanning

The following variables allow configuration of global dependency scanning settings.

Configuring specific analyzers used by dependency scanning

The following variables are used for configuring specific analyzers (used for a specific language/framework).

Other variables

The previous tables are not an exhaustive list of all variables that can be used. They contain all specific GitLab and analyzer variables we support and test. There are many variables, such as environment variables, that you can pass in and they do work. This is a large list, many of which we may be unaware of, and as such is not documented.

For example, to pass the non-GitLab environment variable HTTPS_PROXY to all Dependency Scanning jobs, set it as a CI/CD variable in your .gitlab-ci.yml file like this:

variables:
  HTTPS_PROXY: "https://squid-proxy:3128"

Alternatively we may use it in specific jobs, like Dependency Scanning:

dependency_scanning:
  variables:
    HTTPS_PROXY: $HTTPS_PROXY

As we have not tested all variables you may find some do work and others do not. If one does not work and you need it we suggest submitting a feature request or contributing to the code to enable it to be used.

Using a custom SSL CA certificate authority

You can use the ADDITIONAL_CA_CERT_BUNDLE CI/CD variable to configure a custom SSL CA certificate authority. The ADDITIONAL_CA_CERT_BUNDLE value should contain the text representation of the X.509 PEM public-key certificate. For example, to configure this value in the .gitlab-ci.yml file, use the following:

variables:
  ADDITIONAL_CA_CERT_BUNDLE: |
      -----BEGIN CERTIFICATE-----
      MIIGqTCCBJGgAwIBAgIQI7AVxxVwg2kch4d56XNdDjANBgkqhkiG9w0BAQsFADCB
      ...
      jWgmPqF3vUbZE0EyScetPJquRFRKIesyJuBFMAs=
      -----END CERTIFICATE-----

The ADDITIONAL_CA_CERT_BUNDLE value can also be configured as a custom variable in the UI, either as a file, which requires the path to the certificate, or as a variable, which requires the text representation of the certificate.

Using private Maven repositories

If your private Maven repository requires login credentials, you can use the MAVEN_CLI_OPTS CI/CD variable.

Read more on how to use private Maven repositories.

FIPS-enabled images

Introduced in GitLab 14.10.

GitLab also offers FIPS-enabled Red Hat UBI versions of the Gemnasium images. You can therefore replace standard images with FIPS-enabled images.

Gemnasium scanning jobs automatically use FIPS-enabled image when FIPS mode is enabled in the GitLab instance. (Introduced in GitLab 15.0.)

To manually switch to FIPS-enabled images, set the variable DS_IMAGE_SUFFIX to "-fips".

To ensure compliance with FIPS, the FIPS-enabled image of gemnasium-maven uses the OpenJDK packages for RedHat UBI. As a result, it only supports Java 8, 11, and 17.

Dependency scanning for Gradle projects and auto-remediation for Yarn projects are not supported in FIPS mode.

Interacting with the vulnerabilities

Once a vulnerability is found, you can interact with it. Read more on how to address the vulnerabilities.

Solutions for vulnerabilities

Some vulnerabilities can be fixed by applying the solution that GitLab automatically generates. Read more about the solutions for vulnerabilities.

Security Dashboard

The Security Dashboard is a good place to get an overview of all the security vulnerabilities in your groups, projects and pipelines. Read more about the Security Dashboard.

Vulnerabilities database update

For more information about the vulnerabilities database update, see the maintenance table.

Dependency List

An additional benefit of dependency scanning is the ability to view your project's dependencies and their known vulnerabilities. Read more about the Dependency List.

Reports JSON format

The dependency scanning tool emits a JSON report file. For more information, see the schema for this report.

Here's an example dependency scanning report:

{
  "version": "2.0",
  "vulnerabilities": [
    {
      "id": "51e83874-0ff6-4677-a4c5-249060554eae",
      "category": "dependency_scanning",
      "name": "Regular Expression Denial of Service",
      "message": "Regular Expression Denial of Service in debug",
      "description": "The debug module is vulnerable to regular expression denial of service when untrusted user input is passed into the `o` formatter. It takes around 50k characters to block for 2 seconds making this a low severity issue.",
      "severity": "Unknown",
      "solution": "Upgrade to latest versions.",
      "scanner": {
        "id": "gemnasium",
        "name": "Gemnasium"
      },
      "location": {
        "file": "yarn.lock",
        "dependency": {
          "package": {
            "name": "debug"
          },
          "version": "1.0.5"
        }
      },
      "identifiers": [
        {
          "type": "gemnasium",
          "name": "Gemnasium-37283ed4-0380-40d7-ada7-2d994afcc62a",
          "value": "37283ed4-0380-40d7-ada7-2d994afcc62a",
          "url": "https://deps.sec.gitlab.com/packages/npm/debug/versions/1.0.5/advisories"
        }
      ],
      "links": [
        {
          "url": "https://nodesecurity.io/advisories/534"
        },
        {
          "url": "https://github.com/visionmedia/debug/issues/501"
        },
        {
          "url": "https://github.com/visionmedia/debug/pull/504"
        }
      ]
    },
    {
      "id": "5d681b13-e8fa-4668-957e-8d88f932ddc7",
      "category": "dependency_scanning",
      "name": "Authentication bypass via incorrect DOM traversal and canonicalization",
      "message": "Authentication bypass via incorrect DOM traversal and canonicalization in saml2-js",
      "description": "Some XML DOM traversal and canonicalization APIs may be inconsistent in handling of comments within XML nodes. Incorrect use of these APIs by some SAML libraries results in incorrect parsing of the inner text of XML nodes such that any inner text after the comment is lost prior to cryptographically signing the SAML message. Text after the comment, therefore, has no impact on the signature on the SAML message.\r\n\r\nA remote attacker can modify SAML content for a SAML service provider without invalidating the cryptographic signature, which may allow attackers to bypass primary authentication for the affected SAML service provider.",
      "severity": "Unknown",
      "solution": "Upgrade to fixed version.\r\n",
      "scanner": {
        "id": "gemnasium",
        "name": "Gemnasium"
      },
      "location": {
        "file": "yarn.lock",
        "dependency": {
          "package": {
            "name": "saml2-js"
          },
          "version": "1.5.0"
        }
      },
      "identifiers": [
        {
          "type": "gemnasium",
          "name": "Gemnasium-9952e574-7b5b-46fa-a270-aeb694198a98",
          "value": "9952e574-7b5b-46fa-a270-aeb694198a98",
          "url": "https://deps.sec.gitlab.com/packages/npm/saml2-js/versions/1.5.0/advisories"
        },
        {
          "type": "cve",
          "name": "CVE-2017-11429",
          "value": "CVE-2017-11429",
          "url": "https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2017-11429"
        }
      ],
      "links": [
        {
          "url": "https://github.com/Clever/saml2/commit/3546cb61fd541f219abda364c5b919633609ef3d#diff-af730f9f738de1c9ad87596df3f6de84R279"
        },
        {
          "url": "https://github.com/Clever/saml2/issues/127"
        },
        {
          "url": "https://www.kb.cert.org/vuls/id/475445"
        }
      ]
    }
  ],
  "remediations": [
    {
      "fixes": [
        {
          "id": "5d681b13-e8fa-4668-957e-8d88f932ddc7",
        }
      ],
      "summary": "Upgrade saml2-js",
      "diff": "ZGlmZiAtLWdpdCBhL...OR0d1ZUc2THh3UT09Cg==" // some content is omitted for brevity
    }
  ]
}

CycloneDX Software Bill of Materials

• Introduced in GitLab 14.8 in Beta.
• Generally available in GitLab 15.7.

In addition to the JSON report file, the Gemnasium Dependency Scanning tool outputs a CycloneDX Software Bill of Materials (SBOM) for each supported lock or build file it detects. These CycloneDX SBOMs are named gl-sbom-<package-type>-<package-manager>.cdx.json, and are saved in the same directory as the detected lock or build files.

For example, if your project has the following structure:

.
├── ruby-project/
│   └── Gemfile.lock
├── ruby-project-2/
│   └── Gemfile.lock
├── php-project/
│   └── composer.lock
└── go-project/
    └── go.sum

Then the Gemnasium scanner generates the following CycloneDX SBOMs:

.
├── ruby-project/
│   ├── Gemfile.lock
│   └── gl-sbom-gem-bundler.cdx.json
├── ruby-project-2/
│   ├── Gemfile.lock
│   └── gl-sbom-gem-bundler.cdx.json
├── php-project/
│   ├── composer.lock
│   └── gl-sbom-packagist-composer.cdx.json
└── go-project/
    ├── go.sum
    └── gl-sbom-go-go.cdx.json

You can download CycloneDX SBOMs the same way as other job artifacts.

Merging multiple CycloneDX SBOMs

You can use a CI/CD job to merge multiple CycloneDX SBOMs into a single SBOM. For example:

stages:
  - test
  - merge-cyclonedx-sboms

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

merge cyclonedx sboms:
  stage: merge-cyclonedx-sboms
  image:
    name: cyclonedx/cyclonedx-cli:0.24.2
    entrypoint: [""]
  script:
    - apt-get update && apt-get install -y jq
    - find . -name "gl-sbom-*.cdx.json" -exec cyclonedx merge --output-file gl-sbom-all.cdx.json --input-files "{}" +
    # remove duplicates from merged file. See https://github.com/CycloneDX/cyclonedx-cli/issues/188 for details.
    - |
      jq '. |
      {
        "bomFormat": .bomFormat,
        "specVersion": .specVersion,
        "serialNumber": .serialNumber,
        "version": .version,
        "metadata": {
          "tools": [
            (.metadata.tools | unique[])
          ]
        },
        "components": [
          (.components | unique[])
        ]
      }' "gl-sbom-all.cdx.json" > gl-sbom-all.cdx.json.tmp && mv gl-sbom-all.cdx.json.tmp gl-sbom-all.cdx.json
    # optional: validate the merged sbom
    - cyclonedx validate --input-version v1_4 --input-file gl-sbom-all.cdx.json
  artifacts:
    paths:
      - gl-sbom-all.cdx.json

GitLab uses CycloneDX Properties to store implementation-specific details in the metadata of each CycloneDX SBOM, such as the location of build and lock files. If multiple CycloneDX SBOMs are merged together, this information is removed from the resulting merged file.

Versioning and release process

Check the Release Process documentation.

Contributing to the vulnerability database

To find a vulnerability, you can search the GitLab Advisory Database. You can also submit new vulnerabilities.

Running dependency scanning in an offline environment

For self-managed GitLab instances in an environment with limited, restricted, or intermittent access to external resources through the internet, some adjustments are required for dependency scanning jobs to run successfully. For more information, see Offline environments.

Requirements for offline dependency scanning

Here are the requirements for using dependency scanning in an offline environment:

• GitLab Runner with the docker or kubernetes executor.

• Docker Container Registry with locally available copies of dependency scanning analyzer images.

• If you have a limited access environment you need to allow access, such as using a proxy, to the advisory database: https://gitlab.com/gitlab-org/security-products/gemnasium-db.git. If you are unable to permit access to https://gitlab.com/gitlab-org/security-products/gemnasium-db.git you must host an offline copy of this git repository and set the GEMNASIUM_DB_REMOTE_URL CI/CD variable to the URL of this repository. For more information on configuration variables, see Dependency Scanning.

  This advisory database is constantly being updated, so you must periodically sync your local copy with GitLab.

GitLab Runner has a default pull policy of always, meaning the runner tries to pull Docker images from the GitLab container registry even if a local copy is available. The GitLab Runner pull_policy can be set to if-not-present in an offline environment if you prefer using only locally available Docker images. However, we recommend keeping the pull policy setting to always if not in an offline environment, as this enables the use of updated scanners in your CI/CD pipelines.

Make GitLab dependency scanning analyzer images available inside your Docker registry

For dependency scanning with all supported languages and frameworks, import the following default dependency scanning analyzer images from registry.gitlab.com into your local Docker container registry:

registry.gitlab.com/security-products/gemnasium:3
registry.gitlab.com/security-products/gemnasium-maven:3
registry.gitlab.com/security-products/gemnasium-python:3

The process for importing Docker images into a local offline Docker registry depends on your network security policy. Consult your IT staff to find an accepted and approved process by which external resources can be imported or temporarily accessed. These scanners are periodically updated with new definitions, and you may be able to make occasional updates on your own.

For details on saving and transporting Docker images as a file, see the Docker documentation on docker save, docker load, docker export, and docker import.

Support for Custom Certificate Authorities

Support for custom certificate authorities was introduced in the following versions.

Set dependency scanning CI/CD job variables to use local dependency scanning analyzers

Add the following configuration to your .gitlab-ci.yml file. You must change the value of SECURE_ANALYZERS_PREFIX to refer to your local Docker container registry. You must also change the value of GEMNASIUM_DB_REMOTE_URL to the location of your offline Git copy of the gemnasium-db advisory database:

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

variables:
  SECURE_ANALYZERS_PREFIX: "docker-registry.example.com/analyzers"
  GEMNASIUM_DB_REMOTE_URL: "gitlab.example.com/gemnasium-db.git"

See explanations of the variables above in the configuration section.

Specific settings for languages and package managers

See the following sections for configuring specific languages and package managers.

JavaScript (npm and yarn) projects

Add the following to the variables section of .gitlab-ci.yml:

RETIREJS_JS_ADVISORY_DB: "example.com/jsrepository.json"
RETIREJS_NODE_ADVISORY_DB: "example.com/npmrepository.json"

Ruby (gem) projects

Add the following to the variables section of .gitlab-ci.yml:

BUNDLER_AUDIT_ADVISORY_DB_REF_NAME: "master"
BUNDLER_AUDIT_ADVISORY_DB_URL: "gitlab.example.com/ruby-advisory-db.git"

Python (pip)

If you need to install Python packages before the analyzer runs, you should use pip install --user in the before_script of the scanning job. The --user flag causes project dependencies to be installed in the user directory. If you do not pass the --user option, packages are installed globally, and they are not scanned and don't show up when listing project dependencies.

Python (setuptools)

If you need to install Python packages before the analyzer runs, you should use python setup.py install --user in the before_script of the scanning job. The --user flag causes project dependencies to be installed in the user directory. If you do not pass the --user option, packages are installed globally, and they are not scanned and don't show up when listing project dependencies.

When using self-signed certificates for your private PyPi repository, no extra job configuration (aside from the template .gitlab-ci.yml above) is needed. However, you must update your setup.py to ensure that it can reach your private repository. Here is an example configuration:

1. Update setup.py to create a dependency_links attribute pointing at your private repository for each dependency in the install_requires list:

   install_requires=['pyparsing>=2.0.3'],
   dependency_links=['https://pypi.example.com/simple/pyparsing'],

2. Fetch the certificate from your repository URL and add it to the project:

   printf "\n" | openssl s_client -connect pypi.example.com:443 -servername pypi.example.com | sed -ne '/-BEGIN CERTIFICATE-/,/-END CERTIFICATE-/p' > internal.crt

3. Point setup.py at the newly downloaded certificate:

   import setuptools.ssl_support
   setuptools.ssl_support.cert_paths = ['internal.crt']

Hosting a copy of the gemnasium_db advisory database

The gemnasium_db Git repository is used by gemnasium, gemnasium-maven, and gemnasium-python as the source of vulnerability data. This repository updates at scan time to fetch the latest advisories. However, due to a restricted networking environment, running this update is sometimes not possible. In this case, a user can do one of the following:

• Host a copy of the advisory database
• Use a local clone

Host a copy of the advisory database

If gemnasium-db is not reachable from within the environment, the user can host their own Git copy. Then the analyzer can be instructed to update the database from the user's copy by using GEMNASIUM_DB_REMOTE_URL:

variables:
  GEMNASIUM_DB_REMOTE_URL: https://users-own-copy.example.com/gemnasium-db/.git

...

Use a local clone

If a hosted copy is not possible, then the user can clone gemnasium-db or create an archive before the scan and point the analyzer to the directory (using: GEMNASIUM_DB_LOCAL_PATH). Turn off the analyzer's self-update mechanism (using: GEMNASIUM_DB_UPDATE_DISABLED). In this example, the database directory is created in the before_script, before the gemnasium analyzer's scan job:

...

gemnasium-dependency_scanning:
  variables:
    GEMNASIUM_DB_LOCAL_PATH: ./gemnasium-db-local
    GEMNASIUM_DB_UPDATE_DISABLED: "true"
  before_script:
    - mkdir $GEMNASIUM_DB_LOCAL_PATH
    - tar -xzf gemnasium_db.tar.gz -C $GEMNASIUM_DB_LOCAL_PATH

Warnings

We recommend that you use the most recent version of all containers, and the most recent supported version of all package managers and languages. Using previous versions carries an increased security risk because unsupported versions may no longer benefit from active security reporting and backporting of security fixes.

Python projects

Extra care needs to be taken when using the PIP_EXTRA_INDEX_URL environment variable due to a possible exploit documented by CVE-2018-20225:

An issue was discovered in pip (all versions) because it installs the version with the highest version number, even if the user had intended to obtain a private package from a private index. This only affects use of the PIP_EXTRA_INDEX_URL option, and exploitation requires that the package does not already exist in the public index (and thus the attacker can put the package there with an arbitrary version number).

Troubleshooting

Increase log verbosity

When a job log doesn't contain enough information about a dependency-scanning failure, set SECURE_LOG_LEVEL to debug and check the resulting, more verbose log.

Working around missing support for certain languages or package managers

As noted in the "Supported languages" section some dependency definition files are not yet supported. However, Dependency Scanning can be achieved if the language, a package manager, or a third-party tool can convert the definition file into a supported format.

Generally, the approach is the following:

1. Define a dedicated converter job in your .gitlab-ci.yml file. Use a suitable Docker image, script, or both to facilitate the conversion.
2. Let that job upload the converted, supported file as an artifact.
3. Add dependencies: [<your-converter-job>] to your dependency_scanning job to make use of the converted definitions files.

For example, Poetry projects that only have a pyproject.toml file can generate the poetry.lock file as follows.

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

stages:
  - test

gemnasium-python-dependency_scanning:
  # Work around https://gitlab.com/gitlab-org/gitlab/-/issues/32774
  before_script:
    - pip install "poetry>=1,<2"  # Or via another method: https://python-poetry.org/docs/#installation
    - poetry update --lock # Generates the lock file to be analyzed.

Error response from daemon: error processing tar file: docker-tar: relocation error

This error occurs when the Docker version that runs the dependency scanning job is 19.03.0. Consider updating to Docker 19.03.1 or greater. Older versions are not affected. Read more in this issue.

Getting warning message gl-dependency-scanning-report.json: no matching files

For information on this, see the general Application Security troubleshooting section.

Limitation when using rules:exists

The dependency scanning CI template uses the rules:exists syntax. This directive is limited to 10000 checks and always returns true after reaching this number. Because of this, and depending on the number of files in your repository, a dependency scanning job might be triggered even if the scanner doesn't support your project.

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

For information on this, see the GitLab Secure troubleshooting section.

Import multiple certificates for Java-based projects

The gemnasium-maven analyzer reads the contents of the ADDITIONAL_CA_CERT_BUNDLE variable using keytool, which imports either a single certificate or a certificate chain. Multiple unrelated certificates are ignored and only the first one is imported by keytool.

To add multiple unrelated certificates to the analyzer, you can declare a before_script such as this in the definition of the gemnasium-maven-dependency_scanning job:

gemnasium-maven-dependency_scanning:
  before_script:
    - . $HOME/.bashrc # make the java tools available to the script
    - OIFS="$IFS"; IFS=""; echo $ADDITIONAL_CA_CERT_BUNDLE > multi.pem; IFS="$OIFS" # write ADDITIONAL_CA_CERT_BUNDLE variable to a PEM file
    - csplit -z --digits=2 --prefix=cert multi.pem "/-----END CERTIFICATE-----/+1" "{*}" # split the file into individual certificates
    - for i in `ls cert*`; do keytool -v -importcert -alias "custom-cert-$i" -file $i -trustcacerts -noprompt -storepass changeit -keystore /opt/asdf/installs/java/adoptopenjdk-11.0.7+10.1/lib/security/cacerts 1>/dev/null 2>&1 || true; done # import each certificate using keytool (note the keystore location is related to the Java version being used and should be changed accordingly for other versions)
    - unset ADDITIONAL_CA_CERT_BUNDLE # unset the variable so that the analyzer doesn't duplicate the import

Dependency Scanning job fails with message strconv.ParseUint: parsing "0.0": invalid syntax

Invoking Docker-in-Docker is the likely cause of this error. Docker-in-Docker is:

• Disabled by default in GitLab 13.0 and later.
• Unsupported from GitLab 13.4 and later.

To fix this error, disable Docker-in-Docker for dependency scanning. Individual <analyzer-name>-dependency_scanning jobs are created for each analyzer that runs in your CI/CD pipeline.

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

variables:
  DS_DISABLE_DIND: "true"

Message <file> does not exist in <commit SHA>

When the Location of a dependency in a file is shown, the path in the link goes to a specific Git SHA.

If the lock file that our dependency scanning tools reviewed was cached, however, selecting that link redirects you to the repository root, with the message: <file> does not exist in <commit SHA>.

The lock file is cached during the build phase and passed to the dependency scanning job before the scan occurs. Because the cache is downloaded before the analyzer run occurs, the existence of a lock file in the CI_BUILDS_DIR directory triggers the dependency scanning job.

We recommend committing the lock files, which prevents this warning.

You no longer get the latest Docker image after setting DS_MAJOR_VERSION or DS_ANALYZER_IMAGE

If you have manually set DS_MAJOR_VERSION or DS_ANALYZER_IMAGE for specific reasons, and now must update your configuration to again get the latest patched versions of our analyzers, edit your .gitlab-ci.yml file and either:

• Set your DS_MAJOR_VERSION to match the latest version as seen in our current Dependency Scanning template.

• If you hardcoded the DS_ANALYZER_IMAGE variable directly, change it to match the latest line as found in our current Dependency Scanning template. The line number varies depending on which scanning job you edited.

  For example, currently the gemnasium-maven-dependency_scanning job pulls the latest gemnasium-maven Docker image because DS_ANALYZER_IMAGE is set to "$SECURE_ANALYZERS_PREFIX/gemnasium-maven:$DS_MAJOR_VERSION".

Dependency Scanning of setuptools project fails with use_2to3 is invalid error

Support for 2to3 was removed in setuptools version v58.0.0. Dependency Scanning (running python 3.9) uses setuptools version 58.1.0+, which doesn't support 2to3. Therefore, a setuptools dependency relying on lib2to3 fails with this message:

error in <dependency name> setup command: use_2to3 is invalid

To work around this error, downgrade the analyzer's version of setuptools (for example, v57.5.0):

gemnasium-python-dependency_scanning:
  before_script:
    - pip install setuptools==57.5.0

Dependency Scanning of projects using psycopg2 fails with pg_config executable not found error

Scanning a Python project that depends on psycopg2 can fail with this message:

Error: pg_config executable not found.

psycopg2 depends on the libpq-dev Debian package, which is not installed in the gemnasium-python Docker image. To work around this error, install the libpq-dev package in a before_script:

gemnasium-python-dependency_scanning:
  before_script:
    - apt-get update && apt-get install -y libpq-dev

NoSuchOptionException when using poetry config http-basic with CI_JOB_TOKEN

This error can occur when the automatically generated CI_JOB_TOKEN starts with a hyphen (-). To avoid this error, follow Poetry's configuration advice.

Error: Project has <number> unresolved dependencies

The error message Project has <number> unresolved dependencies indicates a dependency resolution problem caused by your gradle.build or gradle.build.kts file. In the current release, gemnasium-maven cannot continue processing when an unresolved dependency is encountered. However, There is an open issue to allow gemnasium-maven to recover from unresolved dependency errors and produce a dependency graph. Until this issue has been resolved, consult the Gradle dependency resolution docs for details on how to fix your gradle.build file.

Setting build constraints when scanning Go projects

Dependency scanning runs within a linux/amd64 container. As a result, the build list generated for a Go project contains dependencies that are compatible with this environment. If your deployment environment is not linux/amd64, the final list of dependencies might contain additional incompatible modules. The dependency list might also omit modules that are only compatible with your deployment environment. To prevent this issue, you can configure the build process to target the operating system and architecture of the deployment environment by setting the GOOS and GOARCH environment variables of your .gitlab-ci.yml file.

For example:

variables:
  GOOS: "darwin"
  GOARCH: "arm64"

You can also supply build tag constraints by using the GOFLAGS variable:

variables:
  GOFLAGS: "-tags=test_feature"

Dependency Scanning of Go projects returns false positives

The go.sum file contains an entry of every module that was considered while generating the project's build list. Multiple versions of a module are included in the go.sum file, but the MVS algorithm used by go build only selects one. As a result, when dependency scanning uses go.sum, it might report false positives.

To prevent false positives, gemnasium only uses go.sum if it is unable to generate the build list for the Go project. If go.sum is selected, a warning occurs:

[WARN] [Gemnasium] [2022-09-14T20:59:38Z] ▶ Selecting "go.sum" parser for "/test-projects/gitlab-shell/go.sum". False positives may occur. See https://gitlab.com/gitlab-org/gitlab/-/issues/321081.