What happened
Replaying the EC2 metadata SSRF March 2025 campaign
- 1
Probing SSRF-prone sites
Mar 13–25, 2025Requests to websites hosted on EC2 carry the metadata address in one of six parameters. F5 calls the targets insecure sites; no source names a site or confirms that any was vulnerable to SSRF. The activity is tied to no CVE.
- 2
Reading instance metadata
Mar 13–25, 2025The requests ask for four metadata paths, among them the role path /meta-data/iam/security-credentials/admin-role. On an instance that allows IMDSv1 the metadata service answers a plain GET with no token, and the role path returns the role's temporary credentials. F5 reports no retrieved credential.
- 3
Using role credentials
Not reportedIf a response had returned the role's temporary credentials, the holder could present them to AWS from any network. BleepingComputer says the impact could include privilege escalation and S3 access. No source reports that any credential was retrieved or used.
F5 Labs observed exploitation attempts from March 13 to 25, 2025 against EC2-hosted sites. No source reports a retrieved credential, a confirmed victim or a use of credentials.
Where cloud-provider controls fit
A cloud provider cannot fix an SSRF flaw in your site's code, and a WAF rule that blocks the metadata address narrows this campaign without closing the flaw. The control that decides the outcome is the metadata service itself. IMDSv2, required on the instance, refuses the plain request an SSRF flaw can send, so no credential is returned. A declarative policy or a service control policy keeps new instances on IMDSv2 and does not change the instances that already run. Azure, Google Cloud and Oracle Cloud refuse the same request with a required header. If credentials do leak, a service control policy with a VPC condition denies calls made with them from outside the VPC they were issued to (the AWS documentation leaves the result for a call with no VPC endpoint unstated), and a permissions boundary caps what the role can do. F5 reports attempts only, so the sources do not show whether any of these controls was in place or needed.
Could this have been prevented?
In hindsight, not as a whole. 1 control would have refused a step outright, and 5 controls would have limited one. 1 step has 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
- 5
- Contained
- 1
- 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 EC2 metadata SSRF March 2025 campaign
1 · Probing SSRF-prone sites
1 Contained1 Not covered3 on other clouds2 · Reading instance metadata
1 Prevented2 Contained3 on other clouds3 · Using role credentials
2 Contained3 on other cloudsKey facts
- Early scouting was logged on March 13, 2025 from one address, the campaign started on March 15 from another and the activity ended on March 25, per F5 Labs; its page gives the length of the flurry as four days and as six days
- The requests rotated six parameter names, chosen with uniform randomness (dest, file, redirect, target, uri and url), and four metadata paths against the address 169.254.169.254
- The source addresses were several, all in one autonomous system and geolocated to France and Romania; F5 concludes a single actor and names none
- F5 reports exploitation attempts only: no retrieved credential, no confirmed victim and no named site
- F5 says that moving to IMDSv2 fully mitigates metadata exposure through SSRF, because the attacker would also need to supply a session token
- AWS told BleepingComputer that SSRF can expose the role credentials held in the instance metadata service and that it launched metadata protections in 2021 as defense in depth
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.Campaign targets Amazon EC2 instance metadata via SSRFF5 Labs · Apr 8, 2025 · ResearchCampaign dates, SSRF parameters, metadata paths, IMDSv1 relevance, no credential or victim confirmed, attribution reasoning, recommendations
- 2.Hackers target SSRF bugs in EC2-hosted sites to steal AWS credentialsBleepingComputer · Apr 9, 2025 · NewsCampaign summary, AWS statement, IMDSv2 description
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.