How the attack works
AWS Support cases are one of the ways an account owner hears that something is wrong, and one of the ways limits get changed. A principal with Support permissions can do both jobs through the API: ResolveCase closes a case, CreateCase opens one, and AddCommunicationToCase answers in an open case. AWS's Threat Technique Catalog describes the first as a technique of its own. An actor who already holds credentials with permission to view and modify Support cases closes cases, especially security-related ones that were generated automatically, to hide the activity and extend the unauthorized access. Whoever reads the case list afterwards sees resolved cases and no reason to act.
Opening a case serves the other half of an abuse. Raising a sending limit or a service quota often runs through Support, so an attacker who wants more capacity from a stolen account asks for it in the owner's name.
AWS states that all AWS Support API operations are logged by CloudTrail, under the event source support.amazonaws.com, and names CreateCase, DescribeCases and ResolveCase as examples. Its example record for CreateCase shows the service code, category code, severity and issue type in requestParameters and the case ID in responseElements.caseId. In that example the case was opened from the console, and both userAgent and userIdentity.invokedBy hold signin.amazonaws.com. AWS publishes no example record for ResolveCase. No incident report was used for this page: it follows AWS's catalog entry and the Support documentation.
What an attacker needs
- Credentials for a principal with AWS Support permissions, such as support:ResolveCase and support:CreateCase, which administrator policies and the AWS managed support policies include.
- A Support plan that includes the AWS Support API for calls made outside the console.
Impact
A resolved case removes a warning before the owner acts on it, which gives the attacker more time in the account. A case opened by the attacker can lead to higher limits and therefore to more abuse at the owner's cost. The case history also tells the attacker what AWS and the owner already know.
Detection rules
AWS Support Case Created Outside the Console
Detects CreateCase calls to AWS Support that were not made from the AWS console. People usually open cases in the console, and a case opened by a script with stolen credentials is one way to ask for higher limits in the owner's name. AWS's example record for a case opened from the console shows signin.amazonaws.com in userAgent, and the rule excludes that value. The example is the only one AWS publishes, so check what the console writes in your own events and adjust the filter if it differs. The service code, category code and severity of the case are in requestParameters and the case ID is in responseElements.caseId. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.
title: AWS Support Case Created Outside the Console
id: 3a806b3d-fe1e-406a-aff6-0720b776b5e5
status: experimental
description: >-
Detects CreateCase calls to AWS Support that were not made from the AWS
console. People usually open cases in the console, and a case opened by a
script with stolen credentials is one way to ask for higher limits in the
owner's name. AWS's example record for a case opened from the console
shows signin.amazonaws.com in userAgent, and the rule excludes that value.
The example is the only one AWS publishes, so check what the console
writes in your own events and adjust the filter if it differs. The service
code, category code and severity of the case are in requestParameters and
the case ID is in responseElements.caseId.
It also matches attempts that failed (errorCode present): check errorCode,
because an unauthorized attempt is worth seeing.
references:
- https://docs.aws.amazon.com/awssupport/latest/user/logging-using-cloudtrail.html
- https://attack.mitre.org/techniques/T1098/
author: Forestall
date: 2026-11-18
tags:
- attack.persistence
- attack.t1098
logsource:
product: aws
service: cloudtrail
detection:
selection:
eventSource: support.amazonaws.com
eventName: CreateCase
filter_console:
userAgent: signin.amazonaws.com
condition: selection and not filter_console
falsepositives:
- Ticketing integrations and automation that open cases through the AWS Support API, and AWS services that open a case on your behalf. Add those roles to a filter_* selection on userIdentity.arn.
level: low
False positives
- Ticketing integrations and automation that open cases through the AWS Support API, and AWS services that open a case on your behalf. Add those roles to a filter_* selection on userIdentity.arn.
AWS Support Case Resolved
Detects ResolveCase calls to AWS Support. AWS's Threat Technique Catalog lists this event for the closing of support cases by an actor who wants to hide activity, in particular security-related cases that were generated automatically. AWS states that all Support API operations are logged by CloudTrail and names ResolveCase among them. It publishes no example record for this call, so the rule matches the event name only: find the case in the request and check who opened it and what it was about. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.
title: AWS Support Case Resolved
id: c3f953df-9482-49a3-823d-eb5034be57e6
status: experimental
description: >-
Detects ResolveCase calls to AWS Support. AWS's Threat Technique Catalog
lists this event for the closing of support cases by an actor who wants
to hide activity, in particular security-related cases that were
generated automatically. AWS states that all Support API operations are
logged by CloudTrail and names ResolveCase among them. It publishes no
example record for this call, so the rule matches the event name only:
find the case in the request and check who opened it and what it was
about.
It also matches attempts that failed (errorCode present): check errorCode,
because an unauthorized attempt is worth seeing.
references:
- https://docs.aws.amazon.com/awssupport/latest/user/logging-using-cloudtrail.html
- https://aws-samples.github.io/threat-technique-catalog-for-aws/Techniques/T1098.A001.html
- https://docs.aws.amazon.com/awssupport/latest/APIReference/API_ResolveCase.html
- https://attack.mitre.org/techniques/T1098/
author: Forestall
date: 2026-11-18
tags:
- attack.persistence
- attack.t1098
logsource:
product: aws
service: cloudtrail
detection:
selection:
eventSource: support.amazonaws.com
eventName: ResolveCase
condition: selection
falsepositives:
- Engineers and administrators who close their own cases once a question is answered. Add the roles of the people who work with AWS Support to a filter_* selection on userIdentity.arn and alert on every other caller.
level: medium
False positives
- Engineers and administrators who close their own cases once a question is answered. Add the roles of the people who work with AWS Support to a filter_* selection on userIdentity.arn and alert on every other caller.
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.
Support case resolved that nobody in the account opened
Cases that people open leave a CreateCase event in the trail. A resolved case with no matching creation was opened by AWS, for example about suspicious activity, and those are the cases an attacker wants to make disappear.
- Collect
CreateCaseevents fromsupport.amazonaws.comand keep the case IDs fromresponseElements.caseId. - Collect
ResolveCaseevents and read the case ID of the request (the caseId parameter of the API). - Alert when a resolved case ID is not among the created ones, or when the caller differs from the principal that opened the case.
EventsCreateCaseResolveCase
LimitsCases that were opened before the start of your log retention have no creation event either. The case ID of the resolve request follows the API reference, and AWS publishes no example record for it, so confirm the field in your own events first.
Triage
- Identify the caller (userIdentity.arn), the source IP address and the user agent. Is it someone who normally works with AWS Support?
- Find out which case was resolved and who opened it. A case about suspicious activity or exposed credentials that was closed by a principal outside the security team is the pattern to look at.
- For a new case, read requestParameters.serviceCode and requestParameters.categoryCode, and check whether anyone asked for a limit increase.
- Check how old the calling identity is and what else it did, with priority on new users, access keys and policy changes.
Response
Once the attack is confirmed:
- Reopen the case, or open a new one, and tell AWS Support that the case was closed without authorization.
- Read the case and act on what AWS reported in it.
- Contain the principal that resolved or opened the case, and review everything else it did.
- Withdraw limit increases that were requested without approval.
Mitigation
- Grant Support case permissions only to the people and roles that work with AWS Support, and leave them out of workload roles and CI roles.
- Keep the alternate contacts of every account on monitored distribution lists outside the account, so that AWS's notices also arrive by email.
- Send AWS Health events and Support case updates to a channel the security team reads, so that a case does not depend on someone opening the console.
- Review resolved cases as part of incident handling, and reopen any that were closed without an answer from your side.
Frequently asked questions
Why would an attacker close a support case?
AWS's Threat Technique Catalog describes it. An actor who can view and modify Support cases closes them, especially security-related cases that were generated automatically, to hide the activity and extend the unauthorized access. A resolved case looks handled to everyone who checks the list later.
Does CloudTrail show which case was resolved?
AWS states that all Support API operations are logged and names `ResolveCase` among them, but it publishes no example record for that call. Its example for `CreateCase` shows the new case ID in `responseElements.caseId`. Expect the case ID of a resolved case in the request, and confirm the field in your own events before you build on it.
Can I tell console activity from API calls?
In AWS's example record for a case opened from the console, `userAgent` and `userIdentity.invokedBy` both hold `signin.amazonaws.com`. A call from a script or an SDK carries the user agent of that client. One rule on this page uses that difference, and it says that the value comes from AWS's example and should be checked against your own events.
Is every resolved case suspicious?
No. People who work with AWS Support resolve their own cases all the time. The event matters when the caller is not one of them, when the identity is new, or when the case was one that AWS opened about the security of the account.