What happened
Replaying the S3 bucket ransom 2024 intrusion
- 1
Access and first look
January 18, 2024An AWS identity of unknown origin is accepted by the AWS API and calls GetSendQuota against Amazon SES and ListBuckets against Amazon S3. Which credential it is, and how it was obtained, is not stated.
- 2
Checking for a way back
February 5, 2024The identity reads each bucket's versioning, replication, Object Lock, logging, request payment and transfer acceleration settings. Versioning is not enabled.
- 3
Data out, buckets gone
February 5, 2024; when the data was taken is not statedNetwork traffic analysis shows that about 2 GB of the more than 1 TB in the buckets is most likely taken; when, and with which calls, is not stated. The identity deletes every bucket with DeleteBucket, and with versioning off no earlier version survives.
- 4
Ransom and a fake way back
February 5, 2024A ransom note asks for 0.2 BTC within five days and claims the files were downloaded. A decoy recovery binary asks for an AWS key, reads a local file and makes no network connection; where the note and the archive were placed is not stated.
Every S3 bucket of the victim was deleted with versioning off, so the data in them was lost; about 2 GB of more than 1 TB was most likely taken, so the promised restore was not possible. Whether a ransom was paid is not stated.
Where cloud-provider controls fit
Every step of this intrusion ran on one AWS identity whose credential and origin are not stated, so no control here is shown to refuse its calls. A service control policy that keeps your identities to known networks bounds where a credential works, and one that denies the creation of new access keys bounds how many can leak. A service control policy that keeps bucket deletes, object deletes and lifecycle rules to approved administrators bounds who can empty your buckets; service control policies bind only member accounts of an AWS organization, and membership is not stated. S3 Object Lock in compliance mode keeps locked object versions from deletion, but it needs versioning, which was off here, and so does AWS Backup for Amazon S3. The ransom note and the decoy tool sit outside every cloud: the responder found that most likely only about 2 GB was taken, so the promised restore was not possible anyway.
Could this have been prevented?
In hindsight, no control would have stopped it. 8 controls would have limited how far it went. 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.
- 0
- Prevented
- 8
- 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 S3 bucket ransom 2024 intrusion
1 · Access and first look
3 Contained6 on other clouds2 · Checking for a way back
1 Contained3 on other clouds3 · Data out, buckets gone
4 Contained6 on other clouds4 · Ransom and a fake way back
1 Not coveredKey facts
- On January 18, 2024 the attacker's AWS identity called GetSendQuota against Amazon SES and ListBuckets against Amazon S3, the first actions traced in CloudTrail
- On February 5, 2024 it read each bucket's versioning, replication, Object Lock, logging, request payment and transfer acceleration settings; versioning was not enabled
- The same day it deleted all the S3 buckets with DeleteBucket and left a ransom note and a decoy recovery archive
- The note asked for 0.2 BTC within five days and claimed the files were downloaded; the decoy recovery binary asked for an AWS key and made no network connection
- Network traffic analysis showed that the attackers most likely took only about 2 GB of the more than 1 TB in the buckets, so a restore would not have been possible anyway
- Which credential made the calls, how it was obtained, how many buckets were deleted and whether a ransom was paid are not stated; dfir.ch published the case on February 21, 2024 and InfoGuard on February 27, 2024
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.AWS Ransomwaredfir.ch (Stephan Berger) · Feb 21, 2024 · ResearchReconnaissance calls, bucket configuration reads, bucket deletion, versioning state, ransom note, decoy recovery binary, exfiltrated volume, timeline
- 2.AWS ransomware - and what's really behind itInfoGuard · Feb 27, 2024 · ResearchReconnaissance calls, bucket configuration reads, bucket deletion, versioning state, ransom note, decoy recovery binary, exfiltrated volume, responder identity
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.