How the attack works
Amazon SES sends mail for whoever holds credentials with SES permissions, at the cost of the account owner and with the sending reputation of the account. New accounts start in the SES sandbox, where mail can go only to verified addresses and at most 200 messages are sent in 24 hours, and the sandbox status is separate for each AWS Region. An attacker who finds a key therefore first learns where sending is possible. GetSendQuota and GetAccount return the quota and the status of the account in a Region, and ListIdentities shows the verified senders. Where the account is still in the sandbox, PutAccountDetails files the request for production access, and CreateEmailIdentity or VerifyEmailIdentity adds a sender.
Wiz Research described a campaign it identified in May 2025 that followed this order after a first GetCallerIdentity call: quota checks, PutAccountDetails requests that fanned out across all Regions within about ten seconds, an attempt to raise the limits through a support case, and new email identities. The account and identity calls are management events with event source ses.amazonaws.com. The sending itself is harder to see. AWS documents that SendEmail, SendRawEmail and the other sending calls are data events, which a trail does not log unless they are enabled, and that mail sent through the SES SMTP interface is not logged to CloudTrail at all.
What an attacker needs
- Valid credentials for a principal with SES permissions, often a leaked long-term access key with a broad policy.
- For mail to arbitrary recipients, an account that is out of the SES sandbox in at least one Region, or a production access request that AWS grants.
- A verified sender identity, which is one that exists already or one the attacker adds and verifies.
Impact
Phishing and spam sent from the AWS account of the organization, at its cost and with its sending reputation, and from its verified domains if the attacker uses the identities that exist. Bounces and complaints from that mail put the standing of the account as a sender at risk, which also affects legitimate mail. The same credentials usually allow more than SES, so the mail abuse is often the visible part of a wider compromise.
Detection rules
Amazon SES Account Details Changed (Production Access Request)
Detects PutAccountDetails calls to Amazon SES. The call updates the SES account details of the Region it is sent to, and it is how production access is requested: its ProductionAccessEnabled parameter says whether the account should leave the sandbox, in which mail can go only to verified identities. An organization makes this call when it starts sending mail from a Region and rarely afterwards, so a call by a principal that does not manage mail, or in a Region the organization does not send from, is a preparation for sending from the account. The rule matches the event name only, because AWS publishes no example record of this event: read the request parameters and the Region in awsRegion. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.
title: Amazon SES Account Details Changed (Production Access Request)
id: c82df4c9-4fa2-4f3e-bb1a-ae68109a010a
status: experimental
description: >-
Detects PutAccountDetails calls to Amazon SES. The call updates the SES
account details of the Region it is sent to, and it is how production
access is requested: its ProductionAccessEnabled parameter says whether
the account should leave the sandbox, in which mail can go only to
verified identities. An organization makes this call when it starts
sending mail from a Region and rarely afterwards, so a call by a principal
that does not manage mail, or in a Region the organization does not send
from, is a preparation for sending from the account. The rule matches the
event name only, because AWS publishes no example record of this event:
read the request parameters and the Region in awsRegion.
It also matches attempts that failed (errorCode present): check errorCode,
because an unauthorized attempt is worth seeing.
references:
- https://docs.aws.amazon.com/ses/latest/APIReference-V2/API_PutAccountDetails.html
- https://docs.aws.amazon.com/ses/latest/dg/request-production-access.html
- https://docs.aws.amazon.com/ses/latest/dg/logging-using-cloudtrail.html
- https://attack.mitre.org/techniques/T1496/004/
author: Forestall
date: 2026-11-18
tags:
- attack.impact
- attack.t1496.004
logsource:
product: aws
service: cloudtrail
detection:
selection:
eventSource: ses.amazonaws.com
eventName: PutAccountDetails
condition: selection
falsepositives:
- An application team that sets up Amazon SES in a new Region and requests production access, from the console or through infrastructure-as-code. Confirm the request with the team that owns mail.
level: medium
False positives
- An application team that sets up Amazon SES in a new Region and requests production access, from the console or through infrastructure-as-code. Confirm the request with the team that owns mail.
Amazon SES Sender Identity Created
Detects the calls that add a sender identity to Amazon SES: CreateEmailIdentity of the SES v2 API, and VerifyEmailIdentity and VerifyDomainIdentity of the earlier API. AWS describes creating and deleting identities as management events of SES and publishes an example record of VerifyEmailIdentity with the address in requestParameters.emailAddress. An attacker who wants to send from the account adds an address or a domain they control, because mail can be sent only from verified identities. Read the address or domain in the request parameters and check who owns it. For CreateEmailIdentity and VerifyDomainIdentity the rule matches the event name only. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.
title: Amazon SES Sender Identity Created
id: 792136b4-8a2e-40f0-a35d-c9c68872e752
status: experimental
description: >-
Detects the calls that add a sender identity to Amazon SES:
CreateEmailIdentity of the SES v2 API, and VerifyEmailIdentity and
VerifyDomainIdentity of the earlier API. AWS describes creating and
deleting identities as management events of SES and publishes an example
record of VerifyEmailIdentity with the address in
requestParameters.emailAddress. An attacker who wants to send from the
account adds an address or a domain they control, because mail can be
sent only from verified identities. Read the address or domain in the
request parameters and check who owns it. For CreateEmailIdentity and
VerifyDomainIdentity the rule matches the event name only.
It also matches attempts that failed (errorCode present): check errorCode,
because an unauthorized attempt is worth seeing.
references:
- https://docs.aws.amazon.com/ses/latest/dg/logging-using-cloudtrail.html
- https://docs.aws.amazon.com/ses/latest/APIReference-V2/API_CreateEmailIdentity.html
- https://docs.aws.amazon.com/ses/latest/APIReference/API_VerifyDomainIdentity.html
- https://attack.mitre.org/techniques/T1496/004/
author: Forestall
date: 2026-11-18
tags:
- attack.impact
- attack.t1496.004
logsource:
product: aws
service: cloudtrail
detection:
selection:
eventSource: ses.amazonaws.com
eventName:
- CreateEmailIdentity
- VerifyEmailIdentity
- VerifyDomainIdentity
condition: selection
falsepositives:
- Application teams and infrastructure-as-code pipelines that add a sending domain or address for a new service. Add those roles to a filter_* selection on userIdentity.arn and review identities for domains your organization does not own.
level: low
False positives
- Application teams and infrastructure-as-code pipelines that add a sending domain or address for a new service. Add those roles to a filter_* selection on userIdentity.arn and review identities for domains your organization does not own.
Amazon SES Quota or Account Calls in Many Regions by One Principal
Counts, per calling principal (userIdentity.arn), the number of different AWS Regions (awsRegion) in which it read the SES sending quota, the SES account status or the list of identities, or changed the SES account details. Three or more Regions within 10 minutes is the pattern of someone finding out where an account can send mail: the sandbox status of SES is separate for each Region, so the question has to be asked, and production access has to be requested, Region by Region. An application sends from the one or two Regions it is deployed in. Denied calls are counted as well. The count and the window are starting points to tune. The read calls are read-only management events, so the trail has to log Read events and has to cover all Regions. For role sessions the principal is the session ARN, so one role used with different session names is counted per session.
Fires when one userIdentity.arn has at least 3 different awsRegion values within 10 minutes.
title: Amazon SES Quota, Account or Identity Call (Multi-Region Base Event)
id: 627943b5-0294-49ab-87ff-02fadc5c635a
name: aws_ses_quota_account_or_identity_call
status: experimental
description: >-
Base event of the threshold rule in this file. A call to Amazon SES that
reads the sending quota, the account status or the list of identities, or
that changes the account details. Calls that were denied are included.
logsource:
product: aws
service: cloudtrail
detection:
selection:
eventSource: ses.amazonaws.com
eventName:
- GetSendQuota
- GetAccount
- ListIdentities
- ListEmailIdentities
- PutAccountDetails
condition: selection
level: low
---
title: Amazon SES Quota or Account Calls in Many Regions by One Principal
id: 59764689-c5e7-450a-9b55-d2d473264d50
status: experimental
description: >-
Counts, per calling principal (userIdentity.arn), the number of different
AWS Regions (awsRegion) in which it read the SES sending quota, the SES
account status or the list of identities, or changed the SES account
details. Three or more Regions within 10 minutes is the pattern of someone
finding out where an account can send mail: the sandbox status of SES is
separate for each Region, so the question has to be asked, and production
access has to be requested, Region by Region. An application sends from
the one or two Regions it is deployed in. Denied calls are counted as
well. The count and the window are starting points to tune. The read calls
are read-only management events, so the trail has to log Read events and
has to cover all Regions. For role sessions the principal is the session
ARN, so one role used with different session names is counted per session.
references:
- https://docs.aws.amazon.com/ses/latest/dg/request-production-access.html
- https://docs.aws.amazon.com/ses/latest/dg/logging-using-cloudtrail.html
- https://docs.aws.amazon.com/awscloudtrail/latest/userguide/cloudtrail-event-reference-record-contents.html
- https://attack.mitre.org/techniques/T1496/004/
author: Forestall
date: 2026-11-18
tags:
- attack.impact
- attack.t1496.004
correlation:
type: value_count
rules:
- aws_ses_quota_account_or_identity_call
group-by:
- userIdentity.arn
timespan: 10m
condition:
field: awsRegion
gte: 3
falsepositives:
- Security posture scanners and inventory tools that read the SES configuration of every Region on a schedule. Exclude the roles they use.
- Infrastructure-as-code runs that set up SES in several Regions at once.
- The values of 3 Regions and 10 minutes are starting points, not tested limits. Measure what your own tools produce and adjust both.
level: high
False positives
- Security posture scanners and inventory tools that read the SES configuration of every Region on a schedule. Exclude the roles they use.
- Infrastructure-as-code runs that set up SES in several Regions at once.
- The values of 3 Regions and 10 minutes are starting points, not tested limits. Measure what your own tools produce and adjust both.
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.
Quota check, production access request and new identity by one principal
Each step of the abuse is an ordinary SES call. Together and in this order, within an hour and by one principal, they are the preparation for sending from an account that the caller did not set up.
- A
GetSendQuotaorGetAccountevent fromses.amazonaws.com. Note the caller inuserIdentity.arnand itsuserIdentity.accessKeyId. - Within 60 minutes the same caller makes a
PutAccountDetailscall, in the same or in another Region (awsRegion). - Within the same hour the caller adds a sender with
CreateEmailIdentity,VerifyEmailIdentityorVerifyDomainIdentity.
EventsGetSendQuotaGetAccountPutAccountDetailsCreateEmailIdentityVerifyEmailIdentityVerifyDomainIdentity
LimitsThe first setup of SES by an application team follows the same order, although over days and from a known role. The 60 minutes are a starting point. An attacker who finds an account that already has production access skips the request and matches only the first and the last step.
First SES call by an IAM user that never used SES before
Applications that send mail call SES every day with the same identity. A key that has been used for other services for months and now asks SES for quotas and identities has changed hands more often than it has changed purpose.
- Collect the events from
ses.amazonaws.comand keep, for eachuserIdentity.accessKeyIdwhoseuserIdentity.typeisIAMUser, the first time the key called SES. - Alert when a key that has been active in the account for 30 days or more calls SES for the first time.
- Raise the priority when the first calls are
GetSendQuota,GetAccountorListIdentities, and when they come from asourceIPAddressthe key has not used before.
EventsGetSendQuotaGetAccountListIdentities
LimitsA new application feature that starts sending mail looks the same on its first day. The search needs a history of events per access key, and the read calls are recorded only by a trail that logs Read management events. The 30 days are a starting point.
Triage
- Identify the principal (userIdentity.arn), its access key, the source address and the user agent, and decide whether it is one of the applications or people that manage mail.
- List the Regions (awsRegion) in which the principal made SES calls. Calls in Regions where the organization does not send mail are the clearest sign.
- For PutAccountDetails, check whether your organization filed a production access request. For new identities, read the address or domain in the request parameters (VerifyEmailIdentity carries it in requestParameters.emailAddress) and check who owns it.
- Look at what the same access key did before and after, with priority on GetCallerIdentity as a first call, IAM changes and Support cases.
Response
Once the attack is confirmed:
- Deactivate the access key or revoke the role sessions of the principal, and rotate the credential at its source.
- Delete the identities the attacker added, in every Region, and tell AWS Support about production access requests you did not file.
- Check the sending statistics and the bounce and complaint data of the account in each Region, to learn how much mail was sent.
- Review everything else the principal did, because credentials that reach SES usually reach other services as well, and remove what was created.
- Find out how the credential was exposed and close that source, then take the SES permissions away from identities that do not send mail.
Mitigation
- Give SES permissions only to the applications that send mail, and use sending authorization policies to limit who may send from an identity.
- Deny SES with a service control policy in the accounts and Regions that do not send mail.
- Replace long-term access keys with roles and temporary credentials, and remove the keys that are not used.
- Enable CloudTrail data events for SES identities if you rely on CloudTrail to see sending, and watch the sending and reputation metrics of SES.
- Alert on production access requests and new identities, which are rare events in an account whose mail setup is stable.
Frequently asked questions
Why do I not see the phishing mails themselves in CloudTrail?
The sending calls of SES are data events. AWS lists `SendEmail`, `SendRawEmail`, `SendTemplatedEmail` and `SendBulkTemplatedEmail`, and the SES v2 calls `SendEmail` and `SendBulkEmail`, as such, and a trail does not log data events unless you enable them. Mail sent through the SES SMTP interface is not logged to CloudTrail at all. The rules on this page therefore watch the management events around the abuse.
Why does the attacker make the same call in many Regions?
The sandbox status of SES is separate for each Region, and `PutAccountDetails` sets production access for the Region it is sent to. An attacker who wants to know where mail can be sent has to ask each Region, and has to file the request in each Region as well. The threshold rule counts the Regions in which one principal made these calls.
Is a production access request always an attack?
No. An organization that starts sending mail with SES files one for each Region it uses, and it is a rare event after that. A request by a principal that does not manage mail, in a Region you do not use, or in several Regions within seconds is not how an application team works.
References
- MITRE ATT&CK T1496.004: Resource Hijacking: Cloud Service Hijacking
- Logging Amazon SES API calls with AWS CloudTrail (management events, data events, SMTP)
- Request production access (moving out of the Amazon SES sandbox)
- Amazon SES API v2 Reference: PutAccountDetails
- Amazon SES API v2 Reference: CreateEmailIdentity
- Threat Technique Catalog for AWS: Cloud Service Hijacking, SES Messaging (T1496.A001)
- Wiz Research: cloud email abuse campaign that used Amazon SES (4 September 2025)
- Revoke IAM role temporary security credentials