The Gheist campaign is using stolen GitHub tokens to plant malicious workflows that exfiltrate repository secrets, impacting both private repositories and GitHub Enterprise environments since mid-September.
Key Points
- Attackers use stolen tokens to create branches named "ci-sync-" and inject workflows that print repository secrets in base32 format to bypass GitHub's automated masking.
- Security firm Wiz has identified over 13,500 malicious branches across 850 compromised accounts, noting that the campaign deletes logs immediately after exfiltration.
- The Gheist tool targets various credentials, including production environment secrets and, in some cases, attempts to harvest tokens from self-hosted runner machines.
- While GitHub’s automated safeguards block malicious runs in public repositories, these protections do not currently extend to private repositories or GitHub Enterprise Server.
- Wiz recommends that affected organizations rotate all repository secrets, revoke compromised tokens, and audit cloud logs for unauthorized role assumptions.