What was exposed
Replaying the Microsoft December 2019 exposure
- 1
Configured: a network security group change
Dec 5, 2019A change to the database's network security group contains misconfigured security rules that enable exposure of the data; what the rules allow is not stated.
- 2
Reachable: open to the internet with no authentication
Dec 5 to 31, 2019, by Microsoft's datesAnyone with a web browser can read the five Elasticsearch servers with no password; the finder estimates about two days of exposure before the report.
- 3
Found and closed: reported Dec 29, remediated Dec 31
Dec 28 to 31, 2019BinaryEdge indexes the servers, a Comparitech researcher finds them and notifies Microsoft, and Microsoft remediates the configuration on December 31.
Nearly 250 million Customer Service and Support records from 2005 to December 2019, by Comparitech's count, on five Elasticsearch servers that anyone could read with no password or other authentication; most personal information was redacted, but customer email addresses, IP addresses, locations and case notes were in plain text. Open from December 5 to 31 by Microsoft's dates and about two days before the report by the finder's estimate; Microsoft's investigation found no malicious use.
Where cloud-provider controls fit
An Azure Policy Deny on inbound security rules from the internet to database ports refuses a rule that matches it; the sources do not say what the December 5, 2019 rules allowed, so whether it would have refused that change is open. With the rule in place, an Azure Virtual Network Manager security admin rule that denies internet traffic would still have kept readers on the internet out. AWS and Google Cloud bound their own counterparts: an OpenSearch Service domain with VPC access has no public endpoint, and a custom constraint refuses firewall rules from 0.0.0.0/0. An OCI Security Zone removes the internet path instead, and VPC Block Public Access and a hierarchical firewall policy keep internet readers out whatever your project's rules say. Whether the database asks for a password stays with the database software, and the crawl that found it and the fix that closed it stay outside the provider's controls.
Could this have been prevented?
In hindsight, no control would have stopped it. 4 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
- 4
- 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 Microsoft December 2019 exposure
1 · Configured: a network security group change
3 Contained9 on other clouds2 · Reachable: open to the internet with no authentication
1 Contained1 Not covered3 on other clouds3 · Found and closed: reported Dec 29, remediated Dec 31
2 Not coveredKey facts
- On December 5, 2019 a change to the database's network security group contained misconfigured security rules that enabled exposure of the data, per Microsoft; what the rules allowed was not stated
- Per Comparitech, five Elasticsearch servers, each with an apparently identical set, held nearly 250 million Customer Service and Support records from 2005 to December 2019, and no password or other authentication was needed; Microsoft gave no count
- Most personal information had been redacted by automated tools; customer email addresses, IP addresses, locations and case descriptions were in plain text per Comparitech, and data in a non-standard format may have stayed unredacted per Microsoft
- BinaryEdge, a search engine that crawls the internet for exposed data, indexed the databases on December 28; a Comparitech researcher found them on December 29 and notified Microsoft
- Microsoft secured the servers on December 30 to 31 per Comparitech and remediated the configuration on December 31; the data was open from December 5 to 31 by Microsoft's dates, about two days before the report by the finder's estimate and about three weeks per BankInfoSecurity
- Microsoft's investigation found no malicious use; whether anyone else read the data was not stated, and no source reported data taken or used
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.Access Misconfiguration for Customer Support DatabaseMicrosoft Security Response Center (MSRC) · Jan 22, 2020 · Vendor reportDatabase purpose, cause, change date, remediation date, redaction, misuse finding, customer notification, follow-up steps
- 2.Report: 250 million Microsoft customer service and support records exposed on the webComparitech · May 30, 2021 · ResearchRecord count, server type and count, no authentication, record fields, index date, discovery, notification, securing, exposure estimate
- 3.Azure Misconfiguration Exposes 250 Million Microsoft Customer AccountsRCP Magazine (Redmond Channel Partner) · Jan 22, 2020 · NewsAzure hosting, exposure window, notification channel
- 4.Microsoft Exposed 250 Million Customer Support RecordsSecurityWeek · Jan 22, 2020 · NewsSecuring date, other access unknown
- 5.WindiLeaks: 250 million Microsoft customer support records dating back to 2005 exposed to open internetThe Register · Jan 22, 2020 · NewsBinaryEdge role, exposure estimate, securing dates
- 6.Microsoft Error Exposed 250 Million Elasticsearch RecordsInformation Security Media Group (BankInfoSecurity) · Jan 23, 2020 · NewsDiscovery through BinaryEdge, three-week framing of the window
- 7.Azure network security groups overviewMicrosoft Learn (Azure Virtual Network documentation) · Jul 15, 2025 · Vendor reportNetwork security group rules, rule priority, default inbound rules, security admin rules
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.