Skip to main content

Cloud security finding triage

How an AI SOC triages GuardDuty and Security Command Center findings: read the finding, pull the audit trail, and check for an approved change.

Use case
Updated

The short answer

Cloud security finding triage works through alerts from cloud-native tools, such as GuardDuty and Security Command Center, to separate approved changes from threats. Cotool's example GCP Alert Responder, fed by Security Command Center, checks a new service account key against approved changes and records what happened to the key.[1][2]

The problem

Security Command Center can send new findings, finding updates, and findings deleted under retention policies to Pub/Sub, so the stream carries updates as well as new findings.[9]

Except for BigQuery, Data Access audit logs are off by default, so a finding's audit trail may be incomplete until they are enabled.[10]

How an AI SOC handles it

  1. Receive the finding

    Security Command Center findings arrive over Pub/Sub and GuardDuty findings over S3 and SQS. Each create and update is stored with category, severity, resource, and principal.[2][3]

  2. Pull the audit trail

    The agent reads the principal's activity around the finding from gcp_cloud_audit or aws_cloudtrail.[2][3]

  3. Look for an approved change

    The agent searches Jira or Linear for a change ticket that explains the action.[1][4][5]

  4. Check current state

    Through the AWS CLI integration, the agent describes the resource as it is now with a read-only audit policy.[6][11]

  5. Record the outcome

    The verdict names the change or the gap. In Cotool's example, a key created for a CI deployer matched an approved change and was destroyed, so the alert closed as Benign.[1]

Example

A Cloud Logging filter for IAM changestext
logName:"cloudaudit.googleapis.com%2Factivity"
AND protoPayload.methodName=("SetIamPolicy" OR "google.iam.admin.v1.CreateServiceAccountKey")

A filter for IAM policy changes and service account key creation in Admin Activity audit logs, usable in a log sink that feeds Cotool.[2][12]

What people approve

  • Destroying a key or reverting an IAM change is a state-changing action that can require a person's approval.[7]
  • Response agents cannot close an alert as True Positive. When an agent confirms malicious activity, it escalates the alert and a person closes it.[8]

How Cotool handles it

  • Cotool's Respond page lists GCP Security Command Center among the alert sources for its example response agents.[1]

Integrations used

Sources

  1. Respond to every signal (Cotool Respond) · accessed November 16, 2026
  2. GCP Pub/Sub log ingest (Cotool docs) · accessed November 16, 2026
  3. AWS S3/SQS log ingest (Cotool docs) · accessed November 16, 2026
  4. Jira integration (Cotool docs) · accessed November 16, 2026
  5. Linear integration (Cotool docs) · accessed November 16, 2026
  6. Changelog (Cotool docs) · accessed November 16, 2026
  7. AI SOC (Cotool) · accessed November 16, 2026
  8. Alerts (Cotool docs) · accessed November 16, 2026
  9. Enable real-time finding notifications (Security Command Center) · accessed November 16, 2026
  10. Cloud Audit Logs overview (Google Cloud) · accessed November 16, 2026
  11. SecurityAudit managed policy (AWS) · accessed November 16, 2026
  12. Route logs to supported destinations (Cloud Logging docs) · accessed November 16, 2026

Related use cases

All use cases

Related reading

Attackers are scaling with tokens

Cotool helps defenders operate at machine speed. See how security teams are scaling Detection & Response.

Book a demo