CYBERSECURITYTRACKER
Permanent story citation

New Evidence Shows LiteLLM Secret-Stealing Workflows Ran 11 Days Before Supply Chain Attack

This page keeps the story as Cybersecurity Tracker first published it. If the tracker later corrects it, the correction appears below the original and never replaces it.

Back to newsStory 11228

As cited

Copy frozen at (site build).

breaches incidents

New Evidence Shows LiteLLM Secret-Stealing Workflows Ran 11 Days Before Supply Chain Attack

Socket's Threat Research team discovered that malicious secret-exfiltration workflows were executing in the BerriAI/litellm repository as early as October 10, 2026, eleven days before the October 21 supply chain attack that published poisoned litellm==1.107.0 artifacts. The attacker created short-lived GitHub branches, pushed workflows designed to steal CI secrets and transmit them to external collectors, then deleted the branches to minimize detection. Evidence from GitHub Actions performance metrics and environment deployment records shows two successful exfiltration runs: one on October 10 collecting six organization-level secrets including a GitHub token and GitHub App private key, and another on October 12 collecting fourteen additional secrets from environment-scoped deployments, indicating reconnaissance followed by expanded credential harvesting.

Why it matters: Maintainers and organizations hosting packages on PyPI and container registries need to review their supply chain security posture and audit logs to detect similar long-dwell-time compromises, since eleven days of undetected secret theft preceded the public release attack; practitioners should hunt for short-lived branch patterns and unusual GitHub Actions runs that attempt to export secrets.

VendorsGoogleAmazon Web ServicesGitHub
Source published
First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

breaches incidents

New Evidence Shows LiteLLM Secret-Stealing Workflows Ran 11 Days Before Supply Chain Attack

Socket's Threat Research team discovered that malicious secret-exfiltration workflows were executing in the BerriAI/litellm repository as early as October 10, 2026, eleven days before the October 21 supply chain attack that published poisoned litellm==1.107.0 artifacts. The attacker created short-lived GitHub branches, pushed workflows designed to steal CI secrets and transmit them to external collectors, then deleted the branches to minimize detection. Evidence from GitHub Actions performance metrics and environment deployment records shows two successful exfiltration runs: one on October 10 collecting six organization-level secrets including a GitHub token and GitHub App private key, and another on October 12 collecting fourteen additional secrets from environment-scoped deployments, indicating reconnaissance followed by expanded credential harvesting.

Why it matters: Maintainers and organizations hosting packages on PyPI and container registries need to review their supply chain security posture and audit logs to detect similar long-dwell-time compromises, since eleven days of undetected secret theft preceded the public release attack; practitioners should hunt for short-lived branch patterns and unusual GitHub Actions runs that attempt to export secrets.

VendorsAmazon Web ServicesAppleGitHubGoogleMicrosoft
Source published
First seen by Cybersecurity Tracker

Source attribution

Glossary