Technology

Hackers compromise thousands of GitHub accounts to steal tokens and keys

Illustration: Tezkun / AI. Not a photograph of the event.

Attackers used compromised GitHub accounts to inject a malicious script into thousands of repositories, cybersecurity companies Aikido Security, SafeDep, and CodeScoring reported.

According to Aikido Security, credentials of 4,837 compromised accounts were used in the attack. The malicious code, aimed at stealing secrets, was triggered in at least 1,806 projects. SafeDep's analysis recorded changes in more than 8,000 repositories (according to CodeScoring, in 8,926 repositories by the evening of November 12).

The campaign began with test pull requests from the wfx-docs branch. After confirming write permissions, the attackers injected a secrets stealer into GitHub Actions workflows or distributed it via the npm postinstall mechanism. The malware collected GitHub tokens, credentials for AWS, Azure, and Google Cloud services, as well as keys for Kubernetes systems, package registries, and SSH. The gathered information was transmitted to remote servers or to Telegram.

Experts note that the script altered trust policies and added persistent accounts to cloud infrastructure, allowing the attackers to maintain access even after the compromised tokens were revoked. In addition, the attackers attempted to cover their tracks by reverting commits and deleting workflow execution logs.

What it means

Developers and administrators are advised to check commit histories and GitHub Actions run logs starting from November 10 for the presence of wfx-docs branches and "ci: refresh build cache" commits, and if compromised, revoke keys and tokens and review access rules in AWS and Azure clouds.

Published 13 November 2026, 23:01 · No updates · Russian original

Discussion
  1. Loading comments…
No links or insults. One comment per 20 seconds.

Comments on this story live under its post in our Telegram channel. Comment on Telegram