What happened
Replaying the Amazon SES takeover intrusion, May 2025 / no day is given for any call
- 1
Compromised key, SES checked
May 2025; day not statedA compromised long-term access key of an IAM user checks its own identity and reads the account's Amazon SES send quota and account details, to learn whether the account is still in the SES sandbox; how the key was obtained and where the calls came from are not stated.
- 2
Out of the sandbox
May 2025; day not statedWithin ten seconds the key requests SES production access in every AWS region, and AWS support eventually approves one, which lifts the account to the default production quota; a support case for higher limits and an IAM policy that would give the key more rights are both refused for insufficient permissions.
- 3
Sending identities added
May 2025; day not statedThe key adds four attacker-owned domains and legitimate domains with weak DMARC protections as SES sending identities, then sets up role-style sender addresses on them; whether the legitimate domains completed verification is not stated.
- 4
Phishing
Date not statedProofpoint identifies a tax-themed phishing campaign that uses these domains for sending and leads to a credential theft site; whether the mail left through the victim's SES account, and how much was sent, is not stated.
The account's Amazon SES left the sandbox at the default production quota of 50,000 emails a day, and the attacker added its own domains and weakly protected legitimate domains as sending identities; Proofpoint identified a phishing campaign that used those domains. Whether mail was sent through the account is not stated, and the attempts to raise the quota and to give the key more rights failed.
Where the provider's controls fit
On AWS, a service control policy that refuses IAM user calls without MFA would have refused every call the key made after its identity check, from any network, once your applications send with roles rather than IAM user keys. A service control policy that keeps the SES production access request and new sending identities to an approved email administrator role would have kept the account in the sandbox and left the attacker no senders, though AWS support still reviews each request that role makes. A service control policy that denies Amazon SES in accounts that send no email, one that keeps calls to your approved regions and one that keeps calls to your networks could each have narrowed the key's reach further; the write-up names neither the region of the approved request nor where the calls came from. Service control policies on support cases and IAM writes would have refused the two attempts that failed here for insufficient permissions, whatever the user's own policies granted. Each of these needs AWS Organizations, which the write-up does not mention, and none binds the management account, so each caps the reach rather than stopping the step. Where the key leaked is not stated, and no cloud control reaches that place; whether the phishing mail is accepted is decided by each recipient's mail filtering and each domain owner's DMARC policy.
Could this have been prevented?
In hindsight, no control would have stopped it. 11 controls would have limited how far it went. 2 steps have no cloud-provider control. The counts below are control assessments, not attack stages: several can address the same step. The same replay with the controls in place before the incident shows where each phase would have been refused or bounded.
- 0
- Prevented
- 11
- Contained
- 2
- Not covered
How it could have been prevented
The same replay with the controls in place. Green marks where a step is refused, blue where it is bounded, and red where no control reaches it. Open a row to read the control, why it acts, and its caveat.
Replaying the Amazon SES takeover intrusion, May 2025 / no day is given for any call
1 · Compromised key, SES checked
4 Contained1 Not covered3 on other clouds2 · Out of the sandbox
5 Contained3 on other clouds3 · Sending identities added
2 Contained4 · Phishing
1 Not coveredKey facts
- A compromised long-term access key of an IAM user checked its own identity, then read the account's Amazon SES send quota and account details; how the key was obtained was not stated
- Within ten seconds the key sent SES production access requests to every AWS region, and AWS support eventually approved one; the region was not named
- The account kept the default production quota of 50,000 emails a day; a support case to raise the sending limits and an IAM policy to widen the key's rights both failed for insufficient permissions
- The key added four attacker-owned domains and legitimate domains with weak DMARC protections as SES sending identities, then set up role-style sender addresses on them
- Proofpoint identified a tax-themed phishing campaign that used these domains for sending and led to a credential theft site; whether the mail left through the victim's SES account was not stated
- Wiz Research published the write-up on September 4, 2025; it named no victim and described no containment
Sources
Every fact in this breakdown comes from the public reporting below. The replay is a model of that reporting, not a reconstruction of the environment it describes.
- 1.From Compromised Keys to Phishing Campaigns: Inside a Cloud Email Service TakeoverWiz Research · Sep 4, 2025 · ResearchEntry credential, SES reconnaissance, multi-region production access requests and their approval, failed support case, failed IAM policy attempt, sending identities, link to the phishing campaign, recommendations
- 2.Amazon SES Turned Rogue: 50K Phishing Emails a DayeSecurity Planet · Sep 8, 2025 · NewsRe-report of the chain, the sandbox limit as reported
The estate in the replay is illustrative and drawn at the level of attack stages; the victim's real environment is not public. Verdicts describe what a control does when it is in place before the incident. They are not an assessment of the victim's security program.