What happened
Replaying the Uber October 2016 intrusion
- 1
A password opens GitHub
October 2016, day not statedA user's GitHub password opens a private repository for Uber engineers, and its code holds an AWS access key in plain text.
- 2
The AWS key is accepted
From Oct 13, 2016AWS accepts the access key from the repository; its type, owner and permissions are not stated.
- 3
Back-ups downloaded from S3
Oct 13 to Nov 15, 2016With the key the attackers download 16 files of database back-ups from Uber's Amazon S3 datastore.
- 4
Found and closed
Nov 14 to 15, 2016Uber learns of the breach when one of the attackers contacts it, and within 24 hours rotates the AWS credential.
The attackers downloaded 16 files of database back-ups from Uber's Amazon S3 datastore between October 13 and November 15, 2016. Uber says names, email addresses and mobile phone numbers of 57 million users worldwide were downloaded, with the names and driver's license numbers of around 600,000 US drivers. It says no trip location history, card or bank account numbers, Social Security numbers or dates of birth were downloaded, that its corporate systems were not breached, and that it has seen no evidence of fraud or misuse.
Where cloud-provider controls fit
The password and the repository sat in GitHub, outside every cloud provider. The chain met AWS only when an access key copied from Uber's code was presented to the AWS API. On your clouds you can bind a key to your own networks, with a service control policy on AWS, an IAM policy with a network source on Oracle Cloud and Conditional Access for workload identities on Microsoft Entra. You can refuse new long-lived keys, so code written afterwards has none to leak: a service control policy on AWS, an Organization Policy constraint on Google Cloud, an application management policy on Microsoft Entra. You can cap what a stolen key reads in storage with a resource control policy, a permissions boundary and a KMS key policy. Uber learned of the breach from the attackers on November 14, 2016, and says names, email addresses and mobile phone numbers of 57 million users were downloaded.
Could this have been prevented?
In hindsight, no control would have stopped it. 9 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
- 9
- 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 Uber October 2016 intrusion
1 · A password opens GitHub
2 Contained2 Not covered2 · The AWS key is accepted
3 Contained5 on other clouds3 · Back-ups downloaded from S3
4 Contained3 on other clouds4 · Found and closed
1 Not coveredKey facts
- The attackers signed in to GitHub with a password of one user who could reach a private repository for Uber engineers; they said the password was exposed in other large data breaches, which is unconfirmed
- The repository held an AWS access key in plain text in code; the key's type, owner and permissions are not stated
- Between October 13 and November 15, 2016 the attackers downloaded 16 files of database back-ups from Uber's Amazon S3 datastore, per the FTC
- Uber said names, email addresses and mobile phone numbers of 57 million users around the world were downloaded, with the names and driver's license numbers of around 600,000 US drivers
- Uber learned of the breach on November 14, 2016, when one of the attackers contacted it; within 24 hours it implemented multi-factor authentication for GitHub and rotated the AWS credential
- Uber disclosed the incident publicly on November 21, 2017
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.Complaint, In the Matter of Uber Technologies, Inc. (revised complaint, file 152 3054)Federal Trade Commission · Apr 11, 2018 · AdvisoryThe GitHub accounts and the password the intruders said came from other breaches, the plain-text key in a private repository, the Amazon S3 datastore, the 16 files and the October 13 to November 15, 2016 window, the discovery on November 14, 2016, MFA and clear-text storage practices
- 2.Data Security and Bug Bounty Programs: Lessons Learned from the Uber Breach and Security Researchers (S. Hrg. 115-656)US Senate Committee on Commerce, Science, and Transportation (printed record on govinfo.gov) · Feb 6, 2018 · Vendor reportUber's own account: a password of one user, the AWS credential in code on a private repository, the S3 back-up files, the access window, the containment within 24 hours and the security changes since
- 3.2016 Data Security IncidentUber · Nov 21, 2017 · Vendor reportThe disclosure on November 21, 2017, the 57 million users worldwide and around 600,000 US drivers, the data not downloaded, corporate systems not breached, no evidence of misuse
- 4.FTC addresses Uber's undisclosed data breach in new proposed orderFederal Trade Commission (Business Blog) · Apr 12, 2018 · AdvisoryThe order's privacy program, which covers access key management and secure cloud storage
- 5.Uber breach affected 57 million users, covered up for a yearTechTarget (SearchSecurity) · Nov 22, 2017 · NewsBloomberg's report, as relayed, that the private GitHub repository was accessed in October 2016
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.