Skip to content

Web API Fuzz Testing

  • Tier: Ultimate
  • Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated

Web API fuzzing performs fuzz testing of API operation parameters. Fuzz testing sets operation parameters to unexpected values in an effort to cause unexpected behavior and errors in the API backend. This helps you discover bugs and potential security issues that other QA processes may miss.

You should use fuzz testing in addition to the other security scanners in GitLab Secure and your own test processes. If you're using GitLab CI/CD, you can run fuzz tests as part your CI/CD workflow.

For an overview, see Web API Fuzzing.

Getting started

Get started with API fuzzing by editing your CI/CD configuration.

Prerequisites:

To enable API fuzzing:

  • Use the Web API fuzzing configuration form.

    The form lets you choose values for the most common API fuzzing options, and builds a YAML snippet that you can paste in your GitLab CI/CD configuration.

Understanding the results

To view the output of a security scan:

  1. On the left sidebar, select Search or go to and find your project.
  2. On the left sidebar, select Build > Pipelines.
  3. Select the pipeline.
  4. Select the Security tab.
  5. Select a vulnerability to view its details, including:
    • Status: Indicates whether the vulnerability has been triaged or resolved.
    • Description: Explains the cause of the vulnerability, its potential impact, and recommended remediation steps.
    • Severity: Categorized into six levels based on impact. For more information, see severity levels.
    • Scanner: Identifies which analyzer detected the vulnerability.
    • Method: Establishes the vulnerable server interaction type.
    • URL: Shows the location of the vulnerability.
    • Evidence: Describes test case to prove the presence of a given vulnerability
    • Identifiers: A list of references used to classify the vulnerability, such as CWE identifiers.

You can also download the security scan results:

  • In the pipeline's Security tab, select Download results.

For more details, see the pipeline security report.

Findings are generated on feature branches. When they are merged into the default branch, they become vulnerabilities. This distinction is important when evaluating your security posture.

Optimization

To get the most out of API fuzzing, follow these recommendations:

  • Configure runners to use the always pull policy to run the latest versions of the analyzers.

  • By default, API fuzzing downloads all artifacts defined by previous jobs in the pipeline. If your API fuzzing job does not rely on environment_url.txt to define the URL under test or any other files created in previous jobs, you should not download artifacts.

    To avoid downloading artifacts, extend the analyzer CI/CD job to specify no dependencies. For example, for the API fuzzing analyzer, add the following to your .gitlab-ci.yml file:

    apifuzzer_fuzz:
  dependencies: []

Application deployment options

API fuzzing requires a deployed application to be available to scan.

Depending on the complexity of the target application, there are a few options as to how to deploy and configure the API fuzzing template.

Review apps

Review apps are the most involved method of deploying your API Fuzzing target application. To assist in the process, GitLab created a review app deployment using Google Kubernetes Engine (GKE). This example can be found in the Review apps - GKE project, plus detailed instructions to configure review apps in DAST in the README.

Docker Services

If your application uses Docker containers, you have another option for deploying and scanning with API fuzzing. After your Docker build job completes and your image is added to your container registry, you can use the image as a service.

By using service definitions in your .gitlab-ci.yml, you can scan services with the DAST analyzer.

When adding a services section to the job, the alias is used to define the hostname that can be used to access the service. In the following example, the alias: yourapp portion of the dast job definition means that the URL to the deployed application uses yourapp as the hostname (https://yourapp/).

stages:
  - build
  - fuzz

include:
  - template: API-Fuzzing.gitlab-ci.yml

# Deploys the container to the GitLab container registry
deploy:
  services:
  - name: docker:dind
    alias: dind
  image: docker:20.10.16
  stage: build
  script:
    - docker login -u gitlab-ci-token -p $CI_JOB_TOKEN $CI_REGISTRY
    - docker pull $CI_REGISTRY_IMAGE:latest || true
    - docker build --tag $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA --tag $CI_REGISTRY_IMAGE:latest .
    - docker push $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA
    - docker push $CI_REGISTRY_IMAGE:latest

apifuzzer_fuzz:
  services: # use services to link your app container to the dast job
    - name: $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA
      alias: yourapp

variables:
  FUZZAPI_TARGET_URL: https://yourapp

Most applications depend on multiple services such as databases or caching services. By default, services defined in the services fields cannot communicate with each another. To allow communication between services, enable the FF_NETWORK_PER_BUILD feature flag.

variables:
  FF_NETWORK_PER_BUILD: "true" # enable network per build so all services can communicate on the same network

services: # use services to link the container to the dast job
  - name: mongo:latest
    alias: mongo
  - name: $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA
    alias: yourapp

Roll out

Web API fuzzing runs in the fuzz stage of the CI/CD pipeline. To ensure API fuzzing scans the latest code, your CI/CD pipeline should deploy changes to a test environment in one of the stages preceding the fuzz stage.

If your pipeline is configured to deploy to the same web server on each run, running a pipeline while another is still running could cause a race condition in which one pipeline overwrites the code from another. The API to scan should be excluded from changes for the duration of a fuzzing scan. The only changes to the API should be from the fuzzing scanner. Any changes made to the API (for example, by users, scheduled tasks, database changes, code changes, other pipelines, or other scanners) during a scan could cause inaccurate results.

You can run a Web API fuzzing scan using the following methods:

Example API fuzzing projects

Get support or request an improvement

To get support for your particular problem use the getting help channels.

The GitLab issue tracker on GitLab.com is the right place for bugs and feature proposals about API Security and API Fuzzing. Use ~"Category:API Security" label when opening a new issue regarding API fuzzing to ensure it is quickly reviewed by the right people.

Search the issue tracker for similar entries before submitting your own, there's a good chance somebody else had the same issue or feature proposal. Show your support with an emoji reaction or join the discussion.

When experiencing a behavior not working as expected, consider providing contextual information:

  • GitLab version if using a GitLab Self-Managed instance.
  • .gitlab-ci.yml job definition.
  • Full job console output.
  • Scanner log file available as a job artifact named gl-api-security-scanner.log.

Sanitize data attached to a support issue. Remove sensitive information, including: credentials, passwords, tokens, keys, and secrets.

Glossary

  • Assert: Assertions are detection modules used by checks to trigger a fault. Many assertions have configurations. A check can use multiple Assertions. For example, Log Analysis, Response Analysis, and Status Code are common Assertions used together by checks. Checks with multiple Assertions allow them to be turned on and off.
  • Check: Performs a specific type of test, or performed a check for a type of vulnerability. For example, the JSON Fuzzing Check performs fuzz testing of JSON payloads. The API fuzzer is comprised of several checks. Checks can be turned on and off in a profile.
  • Fault: During fuzzing, a failure identified by an Assert is called a fault. Faults are investigated to determine if they are a security vulnerability, a non-security issue, or a false positive. Faults don't have a known vulnerability type until they are investigated. Example vulnerability types are SQL Injection and Denial of Service.
  • Profile: A configuration file has one or more testing profiles, or sub-configurations. You may have a profile for feature branches and another with extra testing for a main branch.