Radware researchers identified 83 malicious GitHub issues, pull requests, or comments across 61 public repositories, including Google LangExtract, Trivy, Brave browser, and MLflow. Most appeared within 43 minutes on September 1, followed by six more in the next two minutes. The researchers suspect one operator used 74 accounts, many with histories suggesting they may have been compromised. The breadth of targets, synchronized timing, and shared infrastructure point to a coordinated effort against widely used open-source projects.
The attack concealed commands in HTML comments, making them invisible in GitHub’s rendered pages. When a repository’s CI/CD workflow responded to an issue or pull-request event, attacker-controlled text could be inserted into shell code and executed, potentially exposing environment variables and sending them to an attacker’s server. Researchers found logs indicating execution in two cases, but could not confirm that credentials were stolen, repository code altered, or downstream software compromised. The likely goal was to obtain credentials and tokens that could enable later changes to trusted projects. Most maintainers were notified, and most affected repositories were closed or fixed, though the researchers could not confirm every case was resolved. They recommend treating external content as untrusted, limiting secrets available to workflows, and separating workflows triggered by outside activity from jobs with elevated permissions.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.