Cloud Logging Best Practices for Security Detection Across AWS, Azure and GCP

· Original article ↗

Summary

The guide recommends prioritizing cloud control-plane logs for threat detection and offers AWS, Azure, GCP and Google Workspace configuration and retention tips, including limitations in Azure read-activity logging.

Key points

  • The framework groups cloud logs by identity, data, network, compute and control-plane use cases; this article focuses on control-plane logs.
  • Control-plane logs can help detect unusual sign-ins, brute-force attempts, resource enumeration, persistence-related creation or modification, and destructive activity.
  • For AWS, the article recommends an Organization Trail to cover new accounts and notes that one copy of ongoing management events delivered to S3 is free.
  • Azure subscription activity logs are retained for 90 days by default, while Entra audit and sign-in log retention ranges from 7 to 30 days depending on license; diagnostic settings can send logs elsewhere for longer retention.
  • The article says Azure control logs do not capture most control-plane read activity, limiting visibility into discovery actions such as listing resources, users and policies.
  • For GCP, Admin Activity logs cover control-plane writes and are enabled by default; Data Access ADMIN_READ logs must be enabled for control-plane reads.
  • Google Workspace logs are enabled by default and retained for six months; streaming them to GCP audit logs can centralize access for security teams.

Article Details

Defense Focus
Improve cloud threat detection and investigation by collecting comprehensive control-plane logs, retaining them adequately, and reviewing sign-in activity for anomalies.
Detection Methods
  • Analyze sign-in logs for sign-ins from unusual countries or providers.
  • Monitor control-plane activity for user, role, virtual machine, and function creation, deletion, or modification.
  • Review logs for password brute force, enumeration across services, unusual resource creation, and data destruction.
  • Enable and monitor read activity where available, including Google Cloud Data Access ADMIN_READ logs for all services.
  • Monitor AWS CloudTrail S3 data events to detect potential data exfiltration targeting S3 buckets.
  • Combine log-based detection with runtime detectors.
Data Sources
  • AWS CloudTrail management events and S3 data events
  • Azure subscription activity logs
  • Microsoft Entra sign-in and audit logs
  • Google Cloud Audit Logs: Admin Activity, Data Access, System Event, and Policy Denied
  • Google Workspace Admin, Login, SAML, OAuth, and Groups logs
Platforms
  • AWS
  • Azure
  • Google Cloud Platform
  • Google Workspace
Defensive Actions
  • Use an AWS Organization trail to cover newly created accounts; deliver management events to an S3 bucket and extend retention or forward logs to another destination as needed.
  • Configure Azure diagnostic settings to retain subscription activity, audit, and sign-in logs beyond their default retention periods.
  • Enable Google Cloud Data Access ADMIN_READ logs for all services to capture control-plane read operations; consider enabling DATA_WRITE where relevant.
  • Stream Google Workspace logs into Google Cloud audit logs for centralized review.
  • Continuously assess logging coverage as the cloud environment changes, including whether new resources are monitored.
  • Use the article's stated defaults and gaps when planning coverage: Azure control logs omit most control-plane read activity, and most Google Workspace log types do not include errors.