What happened
Replaying the AWS intrusion at a SaaS company
- 1
An administrator key from the internet
Dates not statedA leaked access key of an IAM user with AdministratorAccess is used from VPN addresses, scripts list the resources of every region, and the attacker reaches both the staging and the production accounts. How the key leaked and the path between the accounts are not stated.
- 2
Databases opened
Dates not statedThe attacker changes the security groups of the publicly addressed RDS databases to allow its VPN addresses, and resets their master passwords.
- 3
Data out
Dates not statedThe attacker copies S3 bucket content out. Which buckets, what data and how much are not stated.
- 4
Logs, resources and backups destroyed
Dates not statedThe attacker stops and deletes the trail and erases logs from its bucket, turns deletion protection off on EC2 and RDS resources, then deletes the resources and the backups in the same account.
Databases were compromised, S3 data was copied out, the trail and its logs were deleted, and production resources and their backups were deleted; production was down for a week and more than 1,000 developer hours were lost.
Where cloud-provider controls fit
Every call in this intrusion was signed with one long-term IAM user key that held AdministratorAccess, and every production resource, its backups and its trail sat in the organization's management account. A service control policy that refuses IAM user calls without MFA binds only member accounts: it does not bind the management account, and the source does not say which account held the user, so the first control is to keep your workloads out of it. A permissions boundary with a network condition and a bucket policy that admits only your networks also bind the management account, and they refuse calls from addresses outside them. S3 Object Lock keeps the logs written before the trail stops, and AWS Backup Vault Lock in compliance mode keeps the recovery points of a locked vault from deletion by any principal; the source does not say what kind of backups were deleted. Deletion protection on the resources was on, and the key's own rights turned it off.
Could this have been prevented?
In hindsight, not as a whole. 2 controls would have refused a step outright, and 11 controls would have limited one. 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.
- 2
- Prevented
- 11
- Contained
- 0
- 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 AWS intrusion at a SaaS company
1 · An administrator key from the internet
1 Prevented3 Contained9 on other clouds2 · Databases opened
3 Contained6 on other clouds3 · Data out
1 Prevented1 Contained3 on other clouds4 · Logs, resources and backups destroyed
4 Contained9 on other cloudsKey facts
- The entry was the leaked access key of an IAM user that held the AWS managed AdministratorAccess policy; how it leaked and which account held the user were not stated
- Every production resource and its backups sat in one AWS account, which was also the AWS Organizations management account, and the RDS databases had public IPs on public subnets
- The attacker stopped CloudTrail, deleted the trail and deleted logs from its bucket in the same account; GuardDuty raised only a low-severity alert about the stopped trail
- Deletion protection on EC2 and RDS resources was on; the attacker turned it off with the key's own rights and deleted the resources and the backups
- Production was disrupted for a week and more than 1,000 developer hours were lost, according to the consultant's case study
- The case study, published by April 2025, gives no dates for the intrusion and does not name the company
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.Securing a SaaS Company's AWS Environment After a BreachChandrapal Badshah (independent cloud security consultant) · undated · ResearchEntry by a leaked administrator access key, VPN origin, enumeration across regions, security group ingress, RDS master password resets, S3 copy, CloudTrail stopped and trail deleted, deletion protection turned off, resources and backups deleted, staging and production reached, contributing factors, GuardDuty alert, impact, containment and rebuild, recommendations
- 2.AWS breach at a SaaS companyWiz Cloud Threat Landscape · Apr 15, 2025 · ResearchPublication date, summary of the entry vector and impact, threat actor unknown
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.