How the attack works
Switching GuardDuty off is loud. Three quieter changes keep the detector enabled and still remove the findings an attacker cares about. A suppression rule is a saved filter whose action is ARCHIVE: AWS documents that matching findings are archived automatically and are not sent to Security Hub CSPM, Amazon S3, Amazon Detective or Amazon EventBridge, so a SIEM that reads GuardDuty through those paths never sees them. A trusted list (a trusted IP list, called an IPSet in the API, or a trusted entity list) names addresses and domains for which GuardDuty does not generate findings at all, and an entry that appears in both a trusted and a threat list is treated as trusted. A threat list adds findings for known malicious sources, so deleting or deactivating one removes those findings.
Each change is a GuardDuty management event in CloudTrail (guardduty.amazonaws.com): CreateFilter and UpdateFilter with action ARCHIVE in requestParameters, CreateIPSet, UpdateIPSet, CreateTrustedEntitySet and UpdateTrustedEntitySet for trusted lists, and DeleteThreatIntelSet, DeleteThreatEntitySet or an update with activate false for threat lists. In a multi-account setup only the GuardDuty administrator account can make these changes, and the lists it uploads apply to every member account.
What an attacker needs
- Credentials for a principal in the GuardDuty administrator account (or a standalone account) allowed to call guardduty:CreateFilter, UpdateFilter, CreateIPSet, UpdateIPSet, CreateTrustedEntitySet, UpdateTrustedEntitySet or the threat list actions.
- For a trusted list, write access to the Amazon S3 object the list is read from.
Impact
GuardDuty keeps running and looks healthy, but the findings for the attacker's own access keys, roles or addresses are archived before any alerting path sees them, or are never generated. Because administrator-account settings apply to all members, one change can hide the activity across the whole organization. Findings archived by a suppression rule are also not used by Extended Threat Detection to build attack sequences.
Detection rules
GuardDuty Suppression Rule Created or Updated
Detects CreateFilter and UpdateFilter calls whose action is ARCHIVE, which create or change a GuardDuty suppression rule. AWS documents that new findings matching a suppression rule are archived automatically and are not sent to Security Hub CSPM, Amazon S3, Amazon Detective or Amazon EventBridge. It also matches attempts that failed (errorCode present).
title: GuardDuty Suppression Rule Created or Updated
id: a7c8149b-f624-4299-96ed-701ad7fe963f
status: experimental
description: >-
Detects CreateFilter and UpdateFilter calls whose action is ARCHIVE, which
create or change a GuardDuty suppression rule. AWS documents that new
findings matching a suppression rule are archived automatically and are not
sent to Security Hub CSPM, Amazon S3, Amazon Detective or Amazon EventBridge.
It also matches attempts that failed (errorCode present).
references:
- https://docs.aws.amazon.com/guardduty/latest/ug/findings_suppression-rule.html
- https://docs.aws.amazon.com/guardduty/latest/APIReference/API_CreateFilter.html
- https://docs.aws.amazon.com/guardduty/latest/APIReference/API_UpdateFilter.html
author: Forestall
date: 2026-11-18
tags:
- attack.defense-impairment
- attack.t1685
logsource:
product: aws
service: cloudtrail
detection:
selection:
eventSource: guardduty.amazonaws.com
eventName:
- CreateFilter
- UpdateFilter
requestParameters.action: ARCHIVE
condition: selection
falsepositives:
- Suppression rules the security team writes for confirmed false positives, for example a known vulnerability scanner. Confirm the change record and read the filter criteria.
level: medium
False positives
- Suppression rules the security team writes for confirmed false positives, for example a known vulnerability scanner. Confirm the change record and read the filter criteria.
GuardDuty Threat List Deleted or Deactivated
Detects DeleteThreatIntelSet and DeleteThreatEntitySet calls, and UpdateThreatIntelSet and UpdateThreatEntitySet calls with activate false. A threat list makes GuardDuty generate findings for activity that involves known malicious addresses or domains, so removing or deactivating one removes those findings. It also matches attempts that failed (errorCode present).
title: GuardDuty Threat List Deleted or Deactivated
id: 27cc4e43-72c5-4114-8133-226f9ceadb98
status: experimental
description: >-
Detects DeleteThreatIntelSet and DeleteThreatEntitySet calls, and
UpdateThreatIntelSet and UpdateThreatEntitySet calls with activate false.
A threat list makes GuardDuty generate findings for activity that involves
known malicious addresses or domains, so removing or deactivating one removes
those findings. It also matches attempts that failed (errorCode present).
references:
- https://docs.aws.amazon.com/guardduty/latest/ug/guardduty_upload-lists.html
- https://docs.aws.amazon.com/guardduty/latest/APIReference/API_UpdateThreatIntelSet.html
- https://docs.aws.amazon.com/guardduty/latest/APIReference/API_UpdateThreatEntitySet.html
author: Forestall
date: 2026-11-18
tags:
- attack.defense-impairment
- attack.t1685
logsource:
product: aws
service: cloudtrail
detection:
selection_delete:
eventSource: guardduty.amazonaws.com
eventName:
- DeleteThreatIntelSet
- DeleteThreatEntitySet
selection_deactivate:
eventSource: guardduty.amazonaws.com
eventName:
- UpdateThreatIntelSet
- UpdateThreatEntitySet
requestParameters.activate: 'false'
condition: 1 of selection_*
falsepositives:
- Replacing a threat feed with a newer one, or the migration from IP address lists to entity lists that AWS describes. Confirm that a replacement list is active in the same Region.
level: medium
False positives
- Replacing a threat feed with a newer one, or the migration from IP address lists to entity lists that AWS describes. Confirm that a replacement list is active in the same Region.
GuardDuty Trusted IP List or Trusted Entity List Created or Updated
Detects CreateIPSet, UpdateIPSet, CreateTrustedEntitySet and UpdateTrustedEntitySet calls. An IPSet is called a trusted IP list in the console, and AWS documents that GuardDuty does not generate findings for the addresses and domains on a trusted list. An updated list has to be activated again, so a changed list also produces one of these calls. It also matches attempts that failed (errorCode present).
title: GuardDuty Trusted IP List or Trusted Entity List Created or Updated
id: 814c09ce-f542-41e1-b7a1-0f0e44fc6fec
status: experimental
description: >-
Detects CreateIPSet, UpdateIPSet, CreateTrustedEntitySet and
UpdateTrustedEntitySet calls. An IPSet is called a trusted IP list in the
console, and AWS documents that GuardDuty does not generate findings for the
addresses and domains on a trusted list. An updated list has to be activated
again, so a changed list also produces one of these calls. It also matches
attempts that failed (errorCode present).
references:
- https://docs.aws.amazon.com/guardduty/latest/ug/guardduty_upload-lists.html
- https://docs.aws.amazon.com/guardduty/latest/APIReference/API_CreateIPSet.html
- https://docs.aws.amazon.com/guardduty/latest/APIReference/API_UpdateIPSet.html
- https://docs.aws.amazon.com/guardduty/latest/APIReference/API_CreateTrustedEntitySet.html
author: Forestall
date: 2026-11-18
tags:
- attack.defense-impairment
- attack.t1685
logsource:
product: aws
service: cloudtrail
detection:
selection:
eventSource: guardduty.amazonaws.com
eventName:
- CreateIPSet
- UpdateIPSet
- CreateTrustedEntitySet
- UpdateTrustedEntitySet
condition: selection
falsepositives:
- The security team maintaining its trusted list of corporate egress addresses or partner scanners. These changes are rare; confirm each with the team and compare the S3 object with its previous version.
level: high
False positives
- The security team maintaining its trusted list of corporate egress addresses or partner scanners. These changes are rare; confirm each with the team and compare the S3 object with its previous version.
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.
Suppression rule whose criteria name the identity or address that created it
An administrator writes a suppression rule for noise from a known scanner or pipeline. An attacker writes one for the findings their own activity produces, so the filter criteria name the access key, principal or address they are using. A rule whose criteria match the caller of the same call cannot be expressed as a single-event rule, but it is a precise comparison within one CloudTrail record.
- CloudTrail:
CreateFilterorUpdateFilterfromguardduty.amazonaws.comwithactionARCHIVE inrequestParameters. KeeprequestParameters.findingCriteria,userIdentity.arn,userIdentity.accessKeyIdandsourceIPAddress. - Read the criteria for the finding fields that identify a principal or a key, such as
resource.accessKeyDetails.accessKeyId,resource.accessKeyDetails.userIdentity.arnandresource.accessKeyDetails.principalId, and for address criteria. - Alert when a value in the criteria equals the caller's own key, ARN or session issuer, or its
sourceIPAddress, and when the criteria suppress a whole finding type without any other condition.
EventsCreateFilterUpdateFilter
LimitsA security team that creates a rule for its own scanning role produces the same match, so keep a list of approved rules. The criteria are a nested structure, and how a SIEM stores findingCriteria varies; confirm that the field is searchable in your data. An attacker who uses a second identity to write the rule is not matched by the self-reference check, only by the whole-type check.
Triage
- Read the call in CloudTrail. For a filter, look at
findingCriteriainrequestParameters; for a list, atlocation(the S3 object) andactivate. Who made the change (userIdentity.arn), from where (sourceIPAddress) and in which Region (awsRegion)? - Check whether the criteria or the list name the caller's own access key, principal or address, or a broad set such as a whole finding type. A filter that archives findings about the identity that created it is the clearest sign.
- Read the archived findings in the GuardDuty console (Archived view) or through ListFindings with
service.archivedtrue, and the current content of the S3 object behind a trusted list. - Confirm with the security team whether the change was planned. Changes outside the security team's role or pipeline are suspect.
Response
Once the attack is confirmed:
- Delete the filter or set its action back to NOOP, deactivate or restore the trusted list, and re-activate the threat list, in every Region where the call was made.
- Review the findings that were archived while the rule was in place, and the activity of the addresses and principals on the trusted list in CloudTrail, for the whole period.
- Contain the principal that made the change by deactivating its access keys or revoking its role sessions, before it changes the settings again.
- Look for the same principal's other changes to detection, such as GuardDuty detector or member changes and CloudTrail trail changes.
Mitigation
- Keep suppression rules few and narrow, as AWS recommends, and review every rule whose action is ARCHIVE on a schedule.
- Deny guardduty:CreateFilter, UpdateFilter, CreateIPSet, UpdateIPSet, CreateTrustedEntitySet, UpdateTrustedEntitySet, DeleteThreatIntelSet and DeleteThreatEntitySet with an SCP or a permissions boundary except for the security team's role.
- Protect the Amazon S3 objects that hold trusted lists like the GuardDuty configuration itself, with a separate bucket, write access only for the security team, and versioning so a changed list can be compared.
- Prefer suppression rules over trusted lists for known noise. AWS notes that suppressed findings stay visible as archived, while a trusted list prevents the finding from existing.
Frequently asked questions
Are findings lost when a suppression rule archives them?
No. AWS documents that GuardDuty still generates the findings and marks them as archived, keeps them for 90 days, and shows them in the Archived view or through ListFindings with `service.archived` true. They are not sent to Security Hub CSPM, Amazon S3, Detective or EventBridge, which is why a SIEM fed from those paths misses them.
Why is a trusted list more dangerous than a suppression rule?
For an entry on a trusted list GuardDuty does not generate a finding at all, so there is nothing to review afterwards. An entry that is on both a trusted and a threat list is treated as trusted. AWS recommends suppression rules over trusted lists for known noise for that reason.
Do these rules see a change to the file behind an existing trusted list?
Not by itself. The list is read from an Amazon S3 object, and a new upload is an S3 data event. AWS documents that an updated list has to be activated again, which writes `UpdateIPSet` or `UpdateTrustedEntitySet` and is matched here. Log S3 data events on the bucket that holds the lists to see the upload too.
References
- MITRE ATT&CK T1685: Disable or Modify Tools
- Suppression rules in GuardDuty
- Customizing threat detection with entity lists and IP address lists
- Amazon GuardDuty API Reference: CreateFilter
- Amazon GuardDuty API Reference: UpdateFilter
- Amazon GuardDuty API Reference: CreateIPSet
- Amazon GuardDuty API Reference: CreateTrustedEntitySet
- Amazon GuardDuty API Reference: UpdateThreatIntelSet
- Amazon GuardDuty API Reference: UpdateThreatEntitySet
- Logging Amazon GuardDuty API calls with AWS CloudTrail
- AWS re:Post Knowledge Center: Why didn't GuardDuty generate any finding types (suppression rules over trusted IP lists)
- AWS re:Post Knowledge Center: GuardDuty findings for a trusted IP list address (reactivate a changed list)