What was exposed
Replaying the abandoned S3 bucket exposure
- 1
Configured: deleted buckets that are still named
not statedOwners delete S3 buckets that software, documentation and configurations still reference; in the shared global namespace a deleted name can be registered again by any account. When each bucket was deleted, and by whom, is not stated.
- 2
Reachable: any account could take the freed names
about two months before Feb 4, 2025Any account could register the freed names, and watchTowr registered about 150 of them; in about two months the buckets received more than 8 million requests from software on government, military and corporate networks. No source reports a registration of these names by anyone else.
- 3
Found and closed: the names went to AWS
before Feb 4, 2025, dates not statedwatchTowr gives the names to AWS, which blocks those buckets from being re-created and sinkholes them; CISA removes the reference in a 2012 advisory. watchTowr publishes on February 4, 2025; the post describes no file served, and whether anyone else read or registered the names is not stated.
The re-registered buckets received more than 8 million HTTP requests in about two months from software on government, military and corporate networks, among them requests for software updates, unsigned binaries, virtual machine images, CloudFormation templates and SSL VPN configurations. The post describes no file served and the articles report no hostile use; whether anyone else registered or read the names is not stated. AWS blocked those buckets from being re-created and sinkholed them.
Where cloud-provider controls fit
On AWS, a service control policy that denies s3:DeleteBucket except to approved administrators limits who can free a name while something still points to it, and a bucket in the account regional namespace is reserved to its owner even after deletion. A gateway endpoint policy bounds the buckets a workload may reach. Nothing in a cloud account makes a client check a signature or corrects a name written in a document: those stay with the requesting software and the document's owner. No source reports hostile use, and whether anyone else registered or read the names is not stated.
Could this have been prevented?
In hindsight, no control would have stopped it. 3 controls would have limited how far it went. 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.
- 0
- Prevented
- 3
- 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 abandoned S3 bucket exposure
1 · Configured: deleted buckets that are still named
1 Contained3 on other clouds2 · Reachable: any account could take the freed names
2 Contained2 Not covered6 on other clouds3 · Found and closed: the names went to AWS
1 Not coveredKey facts
- watchTowr decided in November 2024 to test the risk and registered about 150 abandoned buckets with logging on, at a cost of "$400+ USD" in the post and $420.85 in The Register
- The buckets received more than 8 million HTTP requests over about two months; CyberScoop says the researchers monitored and took control of the buckets over four months, a disagreement kept here
- Requested files included software updates, unsigned Windows, Linux and macOS binaries, virtual machine images, JavaScript files, CloudFormation templates, SSL VPN configurations and .pom build dependency files; one bucket received over one million requests for .pom files
- The requesting networks, by DNS and WHOIS, included government networks in several countries, military networks, Fortune 100 and Fortune 500 companies, a major payment card network, banks, universities and casinos; the researchers name no bucket and link none to a network
- AWS says the issues occurred when customers deleted buckets that third-party applications still referenced, and that it blocked those specific buckets from being re-created after the names were disclosed; all about 150 were sinkholed
- The post describes no file served and the articles report no hostile use; whether anyone else registered or read the names is not stated
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.
- 1.8 Million Requests Later, We Made The SolarWinds Supply Chain Attack Look AmateurwatchTowr Labs · Feb 4, 2025 · ResearchMethod, bucket count, request counts, requester types, file types, case examples, responses, remediation
- 2.Abandoned AWS S3 buckets can be reused in supply-chain attacks that would make SolarWinds look 'insignificant'The Register · Feb 4, 2025 · NewsCost, AWS statement, sinkholing, no hostile use
- 3.Abandoned Amazon S3 buckets enabled attacks against governments, big firmsSecurityWeek · Feb 5, 2025 · NewsCounts, requester types, AWS statement, hypothetical risk
- 4.Abandoned cloud AWS S3 buckets security risk watchTowrCyberScoop · Feb 4, 2025 · NewsCISA advisory reference, the four-month monitoring period, AWS statement
The estate in the replay is illustrative and drawn at the level of the exposure; the owner's real environment is not public. Verdicts describe what a control does when it is in place before the data is exposed. They are not an assessment of the owner's security program.