Skip to main content
Back
ExposureGoogle Cloud
6 min read

People Data Labs records on an open Google Cloud Elasticsearch server

Owner
Unknown owner (the data came from People Data Labs and OxyData; both deny owning the server)
When
Not stated; first seen October 4, 2019, per news onlyDisclosed Nov 22, 2019
Entry
An Elasticsearch server on a Google Cloud address answered any browser with no authentication; the setting that opened it is not stated.

An Elasticsearch server on a Google Cloud address answered any web browser with no password or authentication of any kind, per the security researchers who found it on October 16, 2019; per the finders it held more than 4 TB, about 4 billion accounts and roughly 1.2 billion unique people, with names, email addresses, phone numbers and social profiles from People Data Labs and OxyData. Both companies said the server was not theirs, and the owner was never identified. One finder notified the FBI and within a few hours someone took the server offline, per an interview; which setting opened it, who closed it and whether anyone else read the data are not stated.

What was exposed

External Parties
Security Researchers
People Data Labs, OxyData
FBI
Internet
Elasticsearch
Enrichment Data Store · Google Cloud
Elasticsearch Server
People Data Labs Indexes
OxyData Indexes

Replaying the PDL Elasticsearch October 2019 exposure

  1. 1

    Configured: the search server answers the internet

    Before Oct 16, 2019; start not stated

    An Elasticsearch server on a Google Cloud address answers requests from the internet with no authentication; the setting that opened it, who set it and when are not stated.

  2. 2

    Reachable: read from a browser with no credential

    Start and closure dates not stated

    Anyone on the internet can read or download all of the server's person records through a web browser, with no credential.

  3. 3

    Found and closed: found on October 16, offline within hours of an FBI notice

    Oct 16, 2019; closure date not stated

    Researchers find the server and check its records against the People Data Labs API; both data companies deny the server, and it goes offline within hours of an FBI notice, per an interview, with the closer not stated.

Impact

More than 4 TB, about 4 billion accounts and roughly 1.2 billion unique people, per the finders: names, email addresses, phone numbers and social profiles copied from People Data Labs and OxyData, readable with no credential from a date not stated until the server went offline within hours of an FBI notice, per an interview; the owner was never identified, and no source reports that anyone other than the finders read the data.

Outcome

Where cloud-provider controls fit

Had the server run on Compute Engine machines in your organization, which the sources do not say, a custom Organization Policy constraint that denies firewall rules open to every internet address would have bounded one way of opening it, and the managed constraint that refuses external IPv4 addresses would have left the machine with no direct path from the internet. For a server in a VPC network under your organization, a hierarchical firewall policy that denies internet ingress to its ports bounds who can read it even where a project rule admits everyone. All three came after the incident. Whether the server asks for a login stays with the Elasticsearch software. The search 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. 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.

External Parties
Security Researchers
People Data Labs, OxyData
FBI
Internet
Elasticsearch
Enrichment Data Store · Google Cloud
Elasticsearch Server
People Data Labs Indexes
OxyData Indexes

Replaying the PDL Elasticsearch October 2019 exposure

1 · Configured: the search server answers the internet

2 Contained3 on other clouds

2 · Reachable: read from a browser with no credential

1 Contained1 Not covered3 on other clouds

3 · Found and closed: found on October 16, offline within hours of an FBI notice

2 Not covered

Key facts

  • An Elasticsearch server at an address hosted with Google Cloud answered any web browser at its public address and port with no password or authentication of any kind, per the finders; the setting that opened it, who set it and when were not stated
  • Per the finders it held more than 4 TB of data and about 4 billion user accounts, among them nearly 3 billion People Data Labs records covering roughly 1.2 billion unique people and about 650 million unique email addresses; Have I Been Pwned loaded 622.2 million unique email addresses from the data
  • The records held names, email addresses, phone numbers and LinkedIn and Facebook profile information; the People Data Labs records also held job positions, short bios and locations, and the OxyData records were mostly scraped LinkedIn profiles
  • The researchers found the server on October 16, 2019; news relayed an earlier first sighting on October 4, 2019 without showing the record
  • The write-up's author looked up their own record with a free People Data Labs account and the API returned what the server held; 50 random users checked the same way gave consistent results
  • People Data Labs and OxyData both said the server was not theirs, and the owner was never identified; one finder notified the FBI and within a few hours someone took the server offline, per an interview, and no source reported that anyone other than the finders read the data

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. 1.
    Personal And Social Information Of 1.2 Billion People Discovered In Massive Data LeakNight Lion Security · Nov 22, 2019 · ResearchGoogle Cloud hosting, Elasticsearch, no authentication, browser access, full download possible, size and counts, indexes, record fields, discovery date, verification against the PDL API, PDL and OxyData denials, owner unknown
  2. 2.
    Unsecured Server Exposed Records of 1.2 Billion: ResearchersInformation Security Media Group (GovInfoSecurity) · Nov 22, 2019 · NewsFBI notification, closure within hours, discovery method, People Data Labs response
  3. 3.
    Account records for up to 1.2B people exposed in massive data leakSiliconANGLE · Nov 22, 2019 · Newsclosure within hours (hedged), FBI investigation (reported), data fields, People Data Labs denial
  4. 4.
    Database with 1.2 billion people's data leaked online without passwordHackRead · Nov 23, 2019 · Newsdiscovery tools (unsourced), FBI report, closure within hours, link to the original post
  5. 5.
    Data-Enriched Profiles on 1.2B People Exposed in Gigantic LeakThreatpost · Nov 22, 2019 · Newsverification attributed to one of the finders, database closed (no date)
  6. 6.
    Public data of 1.2 billion users found on an open Elasticsearch server (Russian)Habr · Nov 23, 2019 · Newsfirst-seen and last-seen dates (relayed), index count, OxyData and People Data Labs responses, counts
  7. 7.
    Data Enrichment Exposure From PDL CustomerHave I Been Pwned · Nov 22, 2019 · Researchunique email addresses loaded, data classes, month listed, owner belief

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.

More cloud attack breakdowns

All breakdowns