
Summary
Detects GitHub Enterprise sign-ins from unrecognized devices or locations by monitoring enterprise audit events. It fires on GitHub audit actions: user.sign_in_from_unrecognized_device, user.sign_in_from_unrecognized_device_and_location, and user.new_device_used, when they appear in logs-github.audit-* for Enterprise Managed User (EMU) environments. A sign-in from an unfamiliar device or location can indicate credential theft or a hijacked session, especially when paired with subsequent changes (new tokens, OAuth grants, SSH keys) that establish persistence. The rule exposes key fields to triage: timestamp, event.action, user.name, github.org, github.repo, github.actor_ip, github.hashed_token, github.programmatic_access_type, github.user_agent, github.device, and github.actor_location.country_name, enabling analysts to assess the legitimacy of the signin and the device, client, and network context. MITRE ATT&CK coverage maps this to T1078 (Valid Accounts) with subtechnique T1078.004 (Cloud Accounts) under Initial Access, reflecting that an unrecognized sign-in could reflect attacker access using stolen credentials.
Investigation steps include: verifying with the user whether a new device was legitimately used; geolocating github.actor_ip and reviewing github.actor_location.country_name and ASN for consistency with the user’s typical activity; distinguishing between user.new_device_used and user.sign_in_from_unrecognized_device_and_location (the latter generally higher risk); correlating with credential changes (personal access tokens, OAuth grants, SSH keys) that followed; and reviewing related events (MFA changes, recovery-code usage) for corroboration.
False positives include legitimate sign-ins from a new corporate device, travel or VPN changes, or IT-provisioned device replacements after proper approval. Remediation guidance includes rotating credentials and tokens, revoking active sessions, re-enrolling MFA, and auditing repository access or changes performed from the unrecognized device; escalate to incident response per playbooks if the IP or device is confirmed malicious. The rule is intended for EMU environments where these events stream to the enterprise audit log and may not appear for personal accounts. The risk score is 47 (medium).
Categories
- Cloud
- Application
Data Sources
- Application Log
ATT&CK Techniques
- T1078
- T1078.004
Created: 2026-09-22