security

Subscribe to all “security” posts via RSS or follow GitHub Changelog on Twitter to stay updated on everything we ship.

→ ~ cd github-changelog
→ ~/github-changelog|main git log main
showing all changes successfully

GitHub now provides a warning about hidden Unicode text

A warning is now displayed when a file’s contents include hidden Unicode text on github.com. Such text can be interpreted differently than it appears in a user interface. For example, hidden Unicode characters can hide text in a file. This can cause code to appear one way and be interpreted another way, especially by AI.

Warning about the presence of hidden Unicode text.

To review a file for which this warning is displayed, open it in an editor that will display the hidden Unicode characters, like Visual Studio Code which highlights the characters by default. Then, verify that the characters are necessary and not disguising text that will be interpreted or compiled differently than it appears.

For more information, refer to Pillar Security: Rules File Backdoor and Hiding and Finding Text with Unicode Tags.

See more

Dependabot adds ARC support for security and version updates in public preview

Actions Runner Controller (ARC) is a Kubernetes operator that automates the deployment, scaling, and lifecycle management of self-hosted actions runners within a Kubernetes cluster. It enables dynamic provisioning of runners based on workflow demand.

Now available in public preview, you can now execute Dependabot on self-hosted GitHub Actions runners managed within a Kubernetes cluster via ARC. This enables auto-scaling, workload isolation, and better resource management for Dependabot jobs, ensuring that dependency updates can run efficiently within an organization’s controlled infrastructure while integrating seamlessly with GitHub Actions.

If you’re running self-hosted runners on Kubernetes with ARC (cloud or on-prem), and you’re interested in trying out Dependabot PR creation, we’d love your feedback.

See more

Credential revocation API to revoke exposed PATs is now generally available

You can now revoke an exposed GitHub personal access token (PAT) you found outside of repositories, even if it’s not yours, to help quickly limit the impact of the exposure and improve the security of the software ecosystem.

If you find classic or fine grained PATs on GitHub or elsewhere, you can submit a bulk revocation request using the new Credential Revocation REST API. If the API receives a valid token, it automatically revokes the token and logs the revocation in the token owner’s audit log. If the exposed token was granted access to a GitHub organization, it will no longer have access to the organization.

A screenshot of the user's audit log event, titled "oauth_access.revoke".

It also notifies the token owner of the revocation through an email sent to the primary email address associated with the owner’s GitHub user account:

A screenshot of an email titled "Action needed: Personal access token was revoked"

This is an unauthenticated API and is available for all users on github.com. To prevent abuse, this API is limited to only 60 unauthenticated requests per hour and a max of 1000 tokens per API request.

Learn more in our documentation on best practices for revoking exposed tokens.

See more

Dependabot reviewers configuration option being replaced by code owners

To improve reliability and reduce review assignment conflicts, we’re removing the Dependabot reviewers configuration option on Tuesday, May 27, 2025.

We’re retiring this dependabot.yml configuration option because the functionality overlaps with GitHub code owners. This has caused issues in the past, and it’s duplicated effort to maintain the same functionality.

Moving forward, we recommend relying solely on code owners for assigning pull request reviewers. You can use a CODEOWNERS file to define individuals or teams responsible for code in a repository. GitHub natively supports code owners, ensuring more consistent and streamlined behavior. This change simplifies your configuration by having all your review requests come from one configuration file.

To get started using code owners today, check out these steps to set up a CODEOWNERS file.

See more

CodeQL adds support for Java 24 and other improvements in version 2.20.6

CodeQL is the static analysis engine behind GitHub code scanning, which finds and remediates security issues in your code. We’ve recently released CodeQL 2.20.6, which brings support for a new version of Java and a variety of other improvements that improve the accuracy of your code scanning results:

Java

  • CodeQL now supports Java version 24
  • We’ve improved the accuracy of the (java/xss) query when javax.servlet.http.HttpServletResponse is used without an exploitable content type

JavaScript / TypeScript

  • We’ve added support for the response threat model, which can be enabled with advanced setup. When enabled, the response data coming back from an outgoing HTTP request is considered a tainted source.
  • We’ve improved the precision of data flow through arrays and call resolution logic, both resulting in improved analysis results

C/C++

  • We’ve improved the accuracy of the cpp/static-buffer-overflow query, resulting in improved results

C#

  • We’ve improved the precision of the cs/call-to-object-tostring query, resulting in improved analysis results

GitHub Actions (Public Preview)

  • We’ve removed the query actions/unversioned-immutable-action from the public suite of queries, which will close any alerts triggered from it

For a full list of changes, please refer to the complete changelog for version 2.20.6. Every new version of CodeQL is automatically deployed to users of GitHub code scanning on github.com. The new functionality in CodeQL 2.20.6 will also be included in GitHub Enterprise Server (GHES) version 3.17. If you use an older version of GHES, you can manually upgrade your CodeQL version.

See more

Dependabot grouped security updates generally available

Dependabot grouped security updates are now generally available. This feature automatically groups Dependabot pull requests, lets you specify several additional options to fine tune your groupings.

You can enable grouped security updates for Dependabot at the repository or organization-level. To enable this feature, go to your repository or organization settings page, then go to the Code security and analysis tab, and click “Enable” for grouped security updates (this also requires each affected repository to enable Dependency graph, Dependabot alerts, and Dependabot security updates). When you enable this feature, Dependabot will collect all available security updates in a repository and attempt to open one pull request with all of them, per ecosystem, across directories.

If you would like more granular control over Dependabot’s grouping, you can also configure the dependabot.yml file in a repository to group by any of the following:

  • Package name
  • Dependency type (production vs development)
  • Semver update level (patch, minor, major)

For additional information, check out the Dependabot configuration file documentation.

For GitHub Enterprise Server users, grouped security updates will be available in Version 3.14.

See more

CodeQL can scan Java projects without a build

With the 2.16.5 release of CodeQL, we’re introducing a new mechanism for creating a CodeQL database for Java codebases, without relying on a build. This enables organizations to more easily adopt CodeQL for Java projects at scale. Note: this release announcement contains details for users of the CodeQL CLI and advanced setup for code scanning. If you’re using GitHub code scanning default setup (which is powered by the CodeQL engine), this related release announcement will likely contain the information you’re looking for.

Previously, CodeQL required a working build to analyze Java projects. This could either be automatically detected or manually specified. Starting with CodeQL 2.16.5, you can now scan Java code without the need for a build. Our large-scale testing has shown that CodeQL can be successfully enabled for over 90% of Java repos without manual intervention.

This feature is currently in public beta and is accessible to all GitHub.com advanced setup for code scanning and CodeQL CLI users scanning Java code:

  • Repositories using advanced setup for code scanning via workflow files will have the option to choose a build-mode. The default value for newly configured Java repos will be build-mode: none.
  • CodeQL CLI users will not experience any change in the default behaviour, for compatibility with existing workflows. Users that want to enable this feature can now use the --build-mode none option. Generally, we also recommend users set the --build-mode option when using the CLI to make it easier to debug and persist the configuration should default behaviour change at any point in the future.
    codeql database create test_no_build_db --language java --build-mode none
  • Repositories containing a mix of Kotlin and Java code still require a working build for CodeQL analysis.

The new mechanism for scanning Java is available on GitHub.com and in CodeQL CLI 2.16.5. While in public beta, this feature will not be available on GitHub Enterprise Server for default setup or advanced setup for code scanning. As we continue to work on scanning Java projects without the need for working builds, send us your feedback.

See more

Java projects no longer require a build when using code scanning via default setup

CodeQL, the static analysis engine that powers GitHub code scanning, can now analyze Java projects without needing a build. This enables organizations to more easily roll out CodeQL at scale. This new way of analyzing Java codebases is now enabled by default for GitHub.com users setting up new repositories with default setup for code scanning.

Previously, CodeQL required a working build to analyze Java projects. This could either be automatically detected or manually specified. By removing that requirement, our large-scale testing has shown that CodeQL can be successfully enabled for over 90% of Java repos without manual intervention.

This feature is currently in public beta and is accessible to all users scanning Java code using default setup for code scanning on GitHub.com:

  • Anyone setting up their repo using code scanning default setup will automatically benefit from this new analysis approach.
  • Repositories containing a mix of Kotlin and Java code still require a working build for CodeQL analysis. CodeQL will default to the autobuild build mode to automatically try and detect the right build command.
  • Repositories with an existing code scanning setup will not experience any changes. If code scanning is working for you today it will continue to work as-is, and there is no need to change your configuration.

GitHub.com users using advanced setup for code scanning and users of the CodeQL CLI will be able to analyze Java projects without needing a working build as part of CodeQL CLI version 2.16.5. While in public beta, this feature will not be available for GitHub Enterprise Server. As we continue to work on scanning Java projects without needing a working build, send us your feedback.

See more

OpenSSF Scorecard info is now available in the Dependency Review Action

Dependency review helps you understand dependency changes and the security impact of these changes at every pull request. We have updated the dependency review action to include information from the OpenSSF Scorecard project into the review, helping you better understand the security posture of the dependencies that you’re using.

See more

Dependabot will now pause scheduled jobs after 15 failures

Previously, if Dependabot encountered 30 consecutive failures, it would stop running scheduled jobs until manual intervention via updating the dependency graph or manifest file. Dependabot will now pause scheduled jobs after 15 failures. This will give an earlier indication of potential issues while still ensuring that critical security updates will continue to be applied without interruption.

Read more in the Dependabot Docs. 

See more

CodeQL 2.16.4: AI-powered autofixes for Java, support for C# 12 & Go 1.22, new queries

CodeQL is the static analysis engine that powers GitHub code scanning. CodeQL version 2.16.4 has been released and has now been rolled out to code scanning users on GitHub.com.

CodeQL code scanning now supports automatic fix suggestions for Java alerts on pull requests, powered by Copilot. This is automatically enabled for all current autofix preview participants. You can sign up for the preview here and use our public discussion for questions and feedback.

The number of generated autofixes is now also visible in a dedicated security overview tile:

security overview showing a counter of fix suggestions

Furthermore, this release

For a full list of changes, please refer to the complete changelog for version 2.16.4. All new functionality will also be included in GHES 3.13. Users of GHES 3.12 or older can upgrade their CodeQL version.

See more

Secret scanning and push protection are enabled by default on new public repositories

All new public repositories owned by personal accounts will now have secret scanning and push protection enabled by default. Pushes to the repository that include known secrets will be blocked by push protection, and any known secrets that are detected in the repository will generate a secret scanning alert. Secret scanning and push protection can be disabled by the repository administrator after the repository is created.

Existing public repositories are not affected, nor are new public repositories that belong to an organization.

See more

CodeQL 2.16.3: AI-powered autofixes for Python, updated queries, and security fixes

CodeQL is the static analysis engine that powers GitHub code scanning. CodeQL version 2.16.3 has been released and has now been rolled out to code scanning users on GitHub.com.

Important changes in this release include:

  • CodeQL code scanning now supports AI-powered automatic fix suggestions for Python alerts on pull requests. This is automatically enabled for all current autofix preview participants.
  • A new option has been added to the Python extractor: python_executable_name. This allows you to select a non-default Python executable installed on the system running the scan (e.g. py.exe on Windows machines).
  • A fix for CVE-2024-25129, a low-severity data exfiltration vulnerability that could be triggered by processing untrusted databases or CodeQL packs.
  • Two new queries:
  • The sinks of queries java/path-injection and java/path-injection-local have been reworked to reduce the number of false positives.

For a full list of changes, please refer to the complete changelog for version 2.16.3. All new functionality will also be included in GHES 3.13. Users of GHES 3.12 or older can upgrade their CodeQL version.

See more
View more changes