What happened
Replaying the Crimson Collective September 2025 intrusion
- 1
Leaked keys confirmed valid
Day not statedLong-term AWS access keys leak from a place no source names, and TruffleHog confirms them with GetCallerIdentity as the first call of every compromised identity.
- 2
An administrator user of its own
Day not statedWith every compromised identity the actor tries to create an IAM user with a console password and access keys, attaches AdministratorAccess to the new user and maps the account.
- 3
Copies of the database and the disks
Day not statedAn RDS master user password is reset and a database snapshot is exported to S3; snapshots of EBS volumes are attached to EC2 instances the actor launched.
- 4
Exfiltration and extortion
Day not statedSelected S3 objects are read out with GetObject, and an extortion note goes out through the victim's Amazon SES and an external email.
Selected S3 objects read out, among them the data the actor had exported, and an extortion note sent through the victim's Amazon SES and an external email; the victims, the volume taken, the ransom and the outcome are not stated.
Where cloud-provider controls fit
A service control policy that refuses new access keys would have kept new long-lived keys out of your accounts, and the existing ones would still need to be found and deleted; an active key still answers the GetCallerIdentity check, since no AWS policy refuses it. Where your keys are meant for known networks, a service control policy with a network condition would have refused the calls from the group's addresses, and a permissions boundary would have capped what each compromised IAM user or role could do. A service control policy that keeps snapshot and launch operations to approved administrators would have bounded the disk copies to identities that are, or can become, an approved administrator, a resource control policy would have refused S3 reads from outside your approved networks, and a deny on Amazon SES sending would have limited who could send through your account. No source says where the keys leaked from, so no control can be placed there, and the external email stays beyond every cloud provider's policy.
Could this have been prevented?
In hindsight, not as a whole. 1 control would have refused a step outright, and 20 controls would have limited one. 3 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.
- 1
- Prevented
- 20
- Contained
- 3
- 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 Crimson Collective September 2025 intrusion
1 · Leaked keys confirmed valid
1 Contained2 Not covered6 on other clouds2 · An administrator user of its own
1 Prevented8 Contained10 on other clouds3 · Copies of the database and the disks
8 Contained15 on other clouds4 · Exfiltration and extortion
3 Contained1 Not covered3 on other cloudsKey facts
- Rapid7 observed the group in two cases in AWS environments in September 2025; no source gives a day for any step, and the victims are unnamed
- CloudTrail showed the TruffleHog user agent as the initial step for every compromised identity; TruffleHog validates a key with GetCallerIdentity, and where the keys leaked from is unconfirmed
- With every compromised identity the actor tried to create an IAM user with a console password and access keys, and it attached the AWS managed AdministratorAccess policy to the new user
- The actor reset an RDS master user password, exported a database snapshot to an S3 bucket whose owner is not stated, and attached EBS snapshots to EC2 instances it launched
- Selected S3 objects were read out with GetObject; the volume taken, the ransom and the outcome are not stated
- The extortion note went out through Amazon SES on the victim's AWS infrastructure and through an external email; Rapid7 listed four IP addresses, the same sets used across the cases
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.Crimson Collective: A New Threat Group Observed Operating in the CloudRapid7 Labs · Oct 7, 2025 · ResearchActor, the two cases, key validation with TruffleHog, IAM persistence and escalation, discovery, RDS and EBS collection, EC2 staging, S3 exfiltration, SES extortion, network indicators, recommendations
- 2.Crimson Collective hackers target AWS cloud instances for data theftBleepingComputer · Oct 8, 2025 · NewsAWS statement on short-term, least-privileged credentials, the extortion note to external email accounts, reporting of the chain
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.