Overview

To set up Datadog SAST in-app, navigate to Security > Code Security.

Select where to run Static Code Analysis scans

Scan with Datadog-hosted scanning

You can run Datadog Static Code Analysis (SAST) scans directly on Datadog infrastructure. Supported repository types include:

• GitHub (excluding repositories that use Git Large File Storage)
• GitLab.com and GitLab Self-Managed
• Azure DevOps

To get started, navigate to the Code Security page.

Scan in CI pipelines

Datadog Static Code Analysis runs in your CI pipelines using the datadog-ci CLI.

First, configure your Datadog API and application keys. Add DD_APP_KEY and DD_API_KEY as secrets. Please ensure your Datadog application key has the code_analysis_read scope.

Next, run Static Code Analysis by following instructions for your chosen CI provider below.

Select your source code management provider

Datadog Static Code Analysis supports all source code management providers, with native support for GitHub, GitLab, and Azure DevOps.

Customize your configuration

By default, Datadog Static Code Analysis (SAST) scans your repositories with Datadog’s default rulesets for each programming language. You can customize which rulesets or rules run, along with other parameters, in Datadog or in a code-security.datadog.yaml file. For the full configuration reference, see Static Code Analysis (SAST) Configuration.

Link findings to Datadog services and teams

Datadog associates code and library scan results with Datadog services and teams to automatically route findings to the appropriate owners. This enables service-level visibility, ownership-based workflows, and faster remediation.

To determine the service where a vulnerability belongs, Datadog evaluates several mapping mechanisms in the order listed in this section.

Each vulnerability is mapped with one method only: if a mapping mechanism succeeds for a particular finding, Datadog does not attempt the remaining mechanisms for that finding.

Mapping using the Software Catalog (recommended)

Services in the Software Catalog identify their codebase content using the codeLocations field. This field is available in the Software Catalog schema version v3 and allows a service to specify:

• a repository URL

apiVersion: v3
kind: service
metadata:
  name: billing-service
  owner: billing-team
datadog:
  codeLocations:
    - repositoryURL: https://github.com/org/myrepo.git

• one or more code paths inside that repository

apiVersion: v3
kind: service
metadata:
  name: billing-service
  owner: billing-team
datadog:
  codeLocations:
    - repositoryURL: https://github.com/org/myrepo.git
      paths:
        - path/to/service/code/**

If you want all the files in a repository to be associated with a service, you can use the glob ** as follows:

apiVersion: v3
kind: service
metadata:
  name: billing-service
  owner: billing-team
datadog:
  codeLocations:
    - repositoryURL: https://github.com/org/myrepo.git
      paths:
        - path/to/service/code/**
    - repositoryURL: https://github.com/org/billing-service.git
      paths:
        - "**"

The schema for this field is described in the Software Catalog entity model.

Datadog goes through all Software Catalog definitions and checks whether the finding’s file path matches. For a finding to be mapped to a service through codeLocations, it must contain a file path.

Some findings might not contain a file path. In those cases, Datadog cannot evaluate codeLocations for that finding, and this mechanism is skipped.

Example (v3 schema)

apiVersion: v3
kind: service
metadata:
  name: billing-service
  owner: billing-team
datadog:
  codeLocations:
    - repositoryURL: https://github.com/org/myrepo.git
      paths:
        - path/to/service/code/**
    - repositoryURL: https://github.com/org/billing-service.git
      paths:
        - "**"

SAST finding

If a vulnerability appeared in github.com/org/myrepo at /src/billing/models/payment.py, then using the codeLocations for billing-service Datadog would add billing-service as an owning service. If your service defines an owner (see above), then Datadog links that team to the finding too. In this case, the finding would be linked to the billing-team.

SCA finding

If a library was declared in github.com/org/myrepo at /go.mod, then Datadog would not match it to billing-service.

Instead, if it was declared in github.com/org/billing-service at /go.mod, then Datadog would match it to billing-service due to the “**” catch-all glob. Consequently, Datadog would link the finding to the billing-team.

When the Software Catalog cannot determine the service

If the Software Catalog does not provide a match, either because the finding’s file path does not match any codeLocations, or because the service uses the v2.x schema, Datadog evaluates whether Error Tracking can identify the service associated with the code. Datadog uses only the last 30 days of Error Tracking data due to product data-retention limits.

When Error Tracking processes stack traces, the traces often include file paths. For example, if an error occurs in: /foo/bar/baz.py, Datadog inspects the directory: /foo/bar. Datadog then checks whether the finding’s file path resides under that directory.

If the finding file is under the same directory:

• Datadog treats this as a strong indication that the vulnerability belongs to the same service.
• The finding inherits the service and team associated with that error in Error Tracking.

If this mapping succeeds, Datadog stops here.

Service inference from file paths or repository names

When neither of the above strategies can determine the service, Datadog inspects naming patterns in the repository and file paths.

Datadog evaluates whether:

• The file path contains identifiers matching a known service.
• The repository name corresponds to a service name.

When using the finding’s file path, Datadog performs a reverse search on each path segment until it finds a matching service or exhausts all options.

For example, if a finding occurs in github.com/org/checkout-service at /foo/bar/baz/main.go, Datadog takes the last path segment, main, and sees if any Software Catalog service uses that name. If there is a match, the finding is attributed to that service. If not, the process continues with baz, then bar, and so on.

When all options have been tried, Datadog checks whether the repository name, checkout-service, matches a Software Catalog service name. If no match is found, Datadog is unsuccessful at linking your finding using Software Catalog.

This mechanism ensures that findings receive meaningful service attribution when no explicit metadata exists.

Link findings to teams through Code Owners

If Datadog is able to link your finding to a service using the above strategies, then the team that owns that service (if defined) is associated with that finding automatically.

Regardless of whether Datadog successfully links a finding to a service (and a Datadog team), Datadog uses the CODEOWNERS information from your finding’s repository to link Datadog and GitHub teams to your findings.

Diff-aware scanning

Diff-aware scanning enables Datadog’s static analyzer to only scan the files modified by a commit in a feature branch. It accelerates scan time significantly by not having the analysis run on every file in the repository for every scan. To enable diff-aware scanning in your CI pipeline, follow these steps:

1. Make sure your DD_APP_KEY, DD_SITE and DD_API_KEY variables are set in your CI pipeline.
2. Add a call to datadog-ci git-metadata upload before invoking the static analyzer. This command ensures that Git metadata is available to the Datadog backend. Git metadata is required to calculate the number of files to analyze.
3. Ensure that the datadog-static-analyzer is invoked with the flag --diff-aware.

Example of commands sequence (these commands must be invoked in your Git repository):

datadog-ci git-metadata upload

datadog-static-analyzer -i /path/to/directory -g -o sarif.json -f sarif –-diff-aware <...other-options...>

Note: When a diff-aware scan cannot be completed, the entire directory is scanned.

Upload third-party static analysis results to Datadog

You can send results from third-party static analysis tools to Datadog, provided they are in the interoperable Static Analysis Results Interchange Format (SARIF) Format. Node.js version 14 or later is required.

To upload a SARIF report:

1. Ensure the DD_API_KEY and DD_APP_KEY variables are defined.

2. Optionally, set a DD_SITE variable (this defaults to datadoghq.com).

3. Install the datadog-ci utility:

   npm install -g @datadog/datadog-ci
   

4. Run the third-party static analysis tool on your code and output the results in the SARIF format.

5. Upload the results to Datadog:

   datadog-ci sarif upload $OUTPUT_LOCATION
   

SARIF Support Guidelines

Datadog supports ingestion of third-party SARIF files that are compliant with the 2.1.0 SARIF schema. The SARIF schema is used differently by static analyzer tools. If you want to send third-party SARIF files to Datadog, please ensure they comply with the following details:

• The violation location is specified through the physicalLocation object of a result.
  • The artifactLocation and it’s uri must be relative to the repository root.
  • The region object is the part of the code highlighted in the Datadog UI.
• The partialFingerprints is used to uniquely identify a finding across a repository.
• properties and tags adds more information:
  • The tag DATADOG_CATEGORY specifies the category of the finding. Acceptable values are SECURITY, PERFORMANCE, CODE_STYLE, BEST_PRACTICES, ERROR_PRONE.
  • The violations annotated with the category SECURITY are surfaced in the Vulnerabilities explorer and the Security tab of the repository view.
• The tool section must have a valid driver section with a name and version attributes.

For example, here’s an example of a SARIF file processed by Datadog:

{
    "runs": [
        {
            "results": [
                {
                    "level": "error",
                    "locations": [
                        {
                            "physicalLocation": {
                                "artifactLocation": {
                                    "uri": "missing_timeout.py"
                                },
                                "region": {
                                    "endColumn": 76,
                                    "endLine": 6,
                                    "startColumn": 25,
                                    "startLine": 6
                                }
                            }
                        }
                    ],
                    "message": {
                        "text": "timeout not defined"
                    },
                    "partialFingerprints": {
                        "DATADOG_FINGERPRINT": "b45eb11285f5e2ae08598cb8e5903c0ad2b3d68eaa864f3a6f17eb4a3b4a25da"
                    },
                    "properties": {
                        "tags": [
                            "DATADOG_CATEGORY:SECURITY",
                            "CWE:1088"
                        ]
                    },
                    "ruleId": "python-security/requests-timeout",
                    "ruleIndex": 0
                }
            ],
            "tool": {
                "driver": {
                    "informationUri": "https://www.datadoghq.com",
                    "name": "<tool-name>",
                    "rules": [
                        {
                            "fullDescription": {
                                "text": "Access to remote resources should always use a timeout and appropriately handle the timeout and recovery. When using `requests.get`, `requests.put`, `requests.patch`, etc. - we should always use a `timeout` as an argument.\n\n#### Learn More\n\n - [CWE-1088 - Synchronous Access of Remote Resource without Timeout](https://cwe.mitre.org/data/definitions/1088.html)\n - [Python Best Practices: always use a timeout with the requests library](https://www.codiga.io/blog/python-requests-timeout/)"
                            },
                            "helpUri": "https://link/to/documentation",
                            "id": "python-security/requests-timeout",
                            "properties": {
                                "tags": [
                                    "CWE:1088"
                                ]
                            },
                            "shortDescription": {
                                "text": "no timeout was given on call to external resource"
                            }
                        }
                    ],
                    "version": "<tool-version>"
                }
            }
        }
    ],
    "version": "2.1.0"
}

SARIF to CVSS severity mapping

The SARIF format defines four severities: none, note, warning, and error. However, Datadog reports violations and vulnerabilities severity using the Common Vulnerability Scoring System (CVSS), which defined five severities: critical, high, medium, low and none.

When ingesting SARIF files, Datadog maps SARIF severities into CVSS severities using the mapping rules below.

Data Retention

Datadog stores findings in accordance with our Data Rentention Periods. Datadog does not store or retain customer source code.