1. Resources
  2. Detection rules
  3. Security Zone or Security Zone Recipe Changed in OCI

Security Zone or Security Zone Recipe Changed in OCI detection rules

An attacker with Cloud Guard management rights weakens or removes a security zone, the guardrail that denies insecure operations in a compartment, so that actions such as making a bucket public are no longer blocked.

Platform
Oracle Cloud Infrastructure
Severity
Medium
Log source
Oracle Cloud Infrastructure Audit
Last updated
MITRE ATT&CK technique
T1685Disable or Modify Tools
Rule formats
Sigmaand queries for: SplunkMicrosoft SentinelIBM QRadar
What an attacker needs
A principal allowed to manage security zones, security zone recipes or the Cloud Guard resource family in the compartment of the zone.
Impact
Operations that the zone used to deny, such as making a bucket public, succeed again in the affected compartments.
Response
Restore the previous recipe, policies, subcompartment or zone first, then use Cloud Guard's policy violations to find what changed during the gap.
Mitigation
Grant manage rights on security zones and recipes to a small security team only, and use the Maximum Security Recipe for production compartments.
API operations
DeleteSecurityZoneRemoveCompartment
Event types
com.oraclecloud.cloudGuard.UpdateSecurityZonecom.oraclecloud.cloudGuard.UpdateSecurityRecipe

How the attack works

A security zone ties one or more compartments to a recipe, a collection of security zone policies. When a resource in the zone is created or changed, Oracle Cloud Infrastructure validates the operation against the policies of the recipe and denies it if one is violated, regardless of which user makes the request. Oracle's own example is a policy that forbids public buckets in Object Storage. The Maximum Security Recipe enables all policies and cannot be modified. Custom recipes can enable and disable individual policies, and Oracle warns that disabled policies can reduce the security posture of the resources in the zone.

A security zone therefore stops an attacker even when the stolen identity has broad IAM permissions. The way around it is to change the zone, and Oracle documents four ways. A zone can be switched to a different recipe. A custom recipe can be edited so that a policy is disabled. A subcompartment can be removed from the zone, after which the policies are no longer enforced on its resources. The zone can be deleted, which Oracle describes as disabling all security zone policies on resources in the zone's compartments. All of them need manage rights on security zones or recipes, or on the Cloud Guard resource family that includes them. No public incident report describes this at the audit event level. The page follows what Oracle documents about security zones and their events.

Oracle's event list names com.oraclecloud.cloudGuard.UpdateSecurityZone and com.oraclecloud.cloudGuard.UpdateSecurityRecipe, and its reference event for security zone recipes carries the recipe name in data.resourceName and the new state, including the list of enabled policies, in data.stateChange.current. The list has no event type for deleting a zone or removing a compartment. The Audit log still records those API calls, and Oracle's Audit reference defines data.eventName as the name of the API operation, so the second rule on this page matches the operation names DeleteSecurityZone and RemoveCompartment. Collection is the same as for the other OCI pages. Ship the _Audit log group of the root compartment (subcompartments included) to the SIEM through a service connector and a stream.

What an attacker needs

  • A principal allowed to manage security-zone or manage security-recipe, or to manage cloud-guard-family, in the compartment of the zone or recipe. Members of Administrators have this.
  • A reason to remove the guardrail, usually an operation in the zone's compartments that a policy of the recipe denies.

Impact

Operations that the zone used to deny succeed again in the affected compartments. Depending on the policies that were dropped, resources can be made public, created without customer-managed encryption, or moved or copied out of the zone. When a zone is deleted, Cloud Guard also stops detecting security zone policy violations for the compartment. The change gives no access by itself, but it removes the control that was meant to hold when an administrator account is compromised.

Detection rules

OCI Security Zone Deleted or Compartment Removed From a Zone

Detects Audit log events whose data.eventName is DeleteSecurityZone or RemoveCompartment. Oracle documents these as the API operations that delete a security zone and remove a subcompartment from one. After a deletion Oracle Cloud Infrastructure no longer enforces security zone policies on the resources of the compartment and Cloud Guard no longer detects policy violations there. After a removal the same holds for the subcompartment. Oracle's list of event types has no entry for these two operations, so the rule matches the operation name, which Oracle's Audit reference defines as the content of data.eventName. Confirm with one test in a non-production compartment that your events carry these names. Read the zone in data.resourceName and the caller in data.identity.principalName.

LevelMediumStatusExperimentalLog sourceOracle Cloud Infrastructure Audit
title: OCI Security Zone Deleted or Compartment Removed From a Zone
id: 193e5d05-adae-478a-baa0-30950f87d6b2
status: experimental
description: >-
  Detects Audit log events whose data.eventName is DeleteSecurityZone or
  RemoveCompartment. Oracle documents these as the API operations that delete
  a security zone and remove a subcompartment from one. After a deletion
  Oracle Cloud Infrastructure no longer enforces security zone policies on
  the resources of the compartment and Cloud Guard no longer detects policy
  violations there. After a removal the same holds for the subcompartment.
  Oracle's list of event types has no entry for these two operations, so the
  rule matches the operation name, which Oracle's Audit reference defines as
  the content of data.eventName. Confirm with one test in a non-production
  compartment that your events carry these names. Read the zone in
  data.resourceName and the caller in data.identity.principalName.
references:
  - https://docs.oracle.com/en-us/iaas/security-zone/using/delete-security-zone.htm
  - https://docs.oracle.com/en-us/iaas/security-zone/using/remove-compartment-from-zone.htm
  - https://docs.oracle.com/en-us/iaas/Content/Audit/Reference/logeventreference.htm
  - https://docs.oracle.com/en-us/iaas/Content/Security/Reference/securityzone_security.htm
  - https://attack.mitre.org/techniques/T1685/
author: Forestall
date: 2026-11-18
tags:
  - attack.defense-impairment
  - attack.t1685
logsource:
  product: oci
  service: audit
detection:
  selection:
    data.eventName:
      - DeleteSecurityZone
      - RemoveCompartment
  condition: selection
falsepositives:
  - Planned restructuring of compartments and security zones by the security team, which is rare. Confirm each event against a change record.
  - Another OCI service that uses the operation name RemoveCompartment for something else. None is known to us. If one appears in your logs, add the event type of the security zone events to the rule.
level: medium

False positives

  • Planned restructuring of compartments and security zones by the security team, which is rare. Confirm each event against a change record.
  • Another OCI service that uses the operation name RemoveCompartment for something else. None is known to us. If one appears in your logs, add the event type of the security zone events to the rule.

OCI Security Zone or Security Zone Recipe Updated

Detects the event types com.oraclecloud.cloudGuard.UpdateSecurityZone and com.oraclecloud.cloudGuard.UpdateSecurityRecipe. Oracle documents that updating a security zone can select a different recipe for it, and that updating a recipe can enable or disable its security zone policies, with the warning that either change can affect the security posture of the resources in the zone. Oracle's reference event for a security zone recipe carries the recipe name in data.resourceName and the state of the recipe, including its list of security policies, in data.stateChange.current. Compare it with data.stateChange.previous to see what changed. An update can also be harmless (a new description) or a hardening step (a policy enabled), so read the change before acting.

LevelMediumStatusExperimentalLog sourceOracle Cloud Infrastructure Audit
title: OCI Security Zone or Security Zone Recipe Updated
id: 1ff3d7ce-8d74-4226-9a40-e3cf50040543
status: experimental
description: >-
  Detects the event types com.oraclecloud.cloudGuard.UpdateSecurityZone and
  com.oraclecloud.cloudGuard.UpdateSecurityRecipe. Oracle documents that
  updating a security zone can select a different recipe for it, and that
  updating a recipe can enable or disable its security zone policies, with
  the warning that either change can affect the security posture of the
  resources in the zone. Oracle's reference event for a security zone recipe
  carries the recipe name in data.resourceName and the state of the recipe,
  including its list of security policies, in data.stateChange.current.
  Compare it with data.stateChange.previous to see what changed. An update
  can also be harmless (a new description) or a hardening step (a policy
  enabled), so read the change before acting.
references:
  - https://docs.oracle.com/en-us/iaas/Content/Events/Reference/eventsproducers.htm
  - https://docs.oracle.com/en-us/iaas/security-zone/using/edit-security-zone.htm
  - https://docs.oracle.com/en-us/iaas/security-zone/using/edit-recipe.htm
  - https://docs.oracle.com/en-us/iaas/Content/Audit/Reference/logeventreference.htm
  - https://attack.mitre.org/techniques/T1685/
author: Forestall
date: 2026-11-18
tags:
  - attack.defense-impairment
  - attack.t1685
logsource:
  product: oci
  service: audit
detection:
  selection:
    eventType:
      - com.oraclecloud.cloudGuard.UpdateSecurityZone
      - com.oraclecloud.cloudGuard.UpdateSecurityRecipe
  condition: selection
falsepositives:
  - The security team maintaining recipes, for example enabling a policy that Oracle added, or editing a description or tags. These objects change rarely, so confirm each event against a change record instead of excluding callers.
level: medium

False positives

  • The security team maintaining recipes, for example enabling a policy that Oracle added, or editing a description or tags. These objects change rarely, so confirm each event against a change record instead of excluding callers.

Detection logic

Patterns that need more than one event, a baseline or knowledge of your own environment. Each is a hypothesis to build and tune in your SIEM, not a ready rule.

Security zone weakened and a previously denied operation performed shortly after

Detection hypothesisSequence of eventsNeeds your own lists

An attacker changes a security zone because it blocks something they want to do. A zone or recipe change that is followed, in the same compartments and by the same principal, by an operation the zone used to deny shows the purpose of the change and separates it from maintenance.

  1. The Audit log records com.oraclecloud.cloudGuard.UpdateSecurityZone or com.oraclecloud.cloudGuard.UpdateSecurityRecipe, or an event whose data.eventName is DeleteSecurityZone or RemoveCompartment. Note data.identity.principalName and the compartments of the zone.
  2. Within 24 hours the same data.identity.principalName performs a write operation on a resource in one of those compartments (data.compartmentId), for example a bucket update.
  3. For a bucket, Oracle's reference event carries data.additionalDetails.publicAccessType. A value other than NoPublicAccess right after the zone change is the clearest case.

EventsUpdateSecurityZoneUpdateSecurityRecipeDeleteSecurityZoneRemoveCompartment

LimitsThe zone events do not list the compartments of the zone, so the reader needs their own map of zones to compartments. Any routine change in the compartment matches the second step, so the logic works best with a short list of operations the dropped policies used to deny. The 24 hours are a starting point.

Triage

  • Identify who made the change (data.identity.principalName) and from where (data.identity.ipAddress), and whether a change record exists.
  • Read which zone or recipe was affected in data.resourceName and data.resourceId, and the compartment in data.compartmentName.
  • For an update, compare data.stateChange.previous with data.stateChange.current to see which recipe the zone uses now or which policies the recipe lost.
  • List the zones that use a changed recipe, because one recipe edit affects every zone it is assigned to.
  • Search the Audit log for what the same principal did in the zone's compartments after the change, such as bucket updates, new public IP addresses, exports or moves of resources.
  • If the change is not approved, restore the recipe or the zone and review the resources that were created or changed while the guardrail was down.

Response

Once the attack is confirmed:

  1. Restore the guardrail first. Assign the previous recipe to the zone again, enable the policies that were disabled, add the removed subcompartment back or create the deleted zone again.
  2. Use the policy violations that Cloud Guard reports for the zone to find resources that were created or changed against the policies while they were not enforced, and correct them.
  3. Contain the principal that made the change. Reset its credentials and remove its rights on security zones and Cloud Guard until its owner is verified.
  4. Review what that principal did in the affected compartments during the gap, with priority on public access, exported data and resources moved to other compartments.
  5. Confirm afterwards that every compartment that should be in a security zone is in one again and that the zone uses the intended recipe.

Mitigation

  • Grant manage on security zones, recipes and cloud-guard-family to a small security team only. Oracle recommends giving delete permissions to a minimum set of users and calls this especially critical for security zones.
  • Use the Maximum Security Recipe, which cannot be edited, for compartments that hold production data, and document every zone that uses a custom recipe and why.
  • Review the recipes and the compartments of each security zone on a schedule, and compare the enabled policies with the Maximum Security Recipe.
  • Alert on every change to a security zone or recipe, since these objects change rarely in a stable tenancy.

Frequently asked questions

What is the difference between a security zone policy and an IAM policy?

Oracle describes it this way. An IAM policy grants users the ability to manage resources in a compartment. A security zone policy denies certain actions, grants nothing, and is validated regardless of which user performs the operation. That is why a security zone still blocks an insecure change when the account making it is an administrator, and why an attacker has to change the zone first.

Does the update rule show which policy was disabled?

The rule matches the event types only. Oracle's reference event for a security zone recipe carries the state of the recipe in `data.stateChange.current`, including its list of security policies, and the Audit reference describes `data.stateChange.previous` as the state before the operation. Compare the two in triage. Oracle publishes a sample only for the creation of a recipe, so check that your update events carry both states.

Why does the second rule match an operation name and not an event type?

Oracle's list of event types names the creation and update of security zones and recipes, and has no entry for deleting a zone or removing a compartment from it. The Audit service records every API call, and Oracle defines `data.eventName` as the name of the API operation that generated the event. The rule therefore matches the documented operation names `DeleteSecurityZone` and `RemoveCompartment` in that field.

Is disabling Cloud Guard itself covered by this page?

No. Oracle documents a Cloud Guard status event for the Events service, with a status value for a disabled Cloud Guard, but it is a notification that Cloud Guard emits and not an Audit log event type on Oracle's list. The rules here read the Audit log, so they cover security zones and recipes only.

References

We respect your privacy

We use cookies to keep this site secure and working properly. With your permission, we also use optional cookies to understand usage and improve the experience. Cookie Policy

You can change your choice at any time.