Demonstration dataset · Synthetic records · Not real security assessments
COLDGUARDResearch Observatory

Attackers

Who is behind the documented disruptions?

Usually, the public record does not say. COLDGUARD reports attribution as a confidence level per finding and groups findings by the adversary-driven pattern they describe. It never names a threat actor or group.

How often attribution is established

Dataset demo-2026.11.07 · methodology 0.1-draft

Cause establishedCause not established
n = 97 published findings; dataset demo-2026.11.07. Each published finding is placed once on the attribution spectrum.
LevelFindings
High attribution42
Moderate attribution34
Limited attribution1
Not attributed20
n = 97 published findings; dataset demo-2026.11.07. Each published finding is placed once on the attribution spectrum.

High attribution

42

43.3% of published findings

Moderate attribution

34

35.1% of published findings

Limited attribution

1

1% of published findings

Not attributed

20

20.6% of published findings

Read together: 76 of 97 published findings carry high or moderate attribution, and 20 carry none. The gap is the headline. It reflects what organizations, regulators, and reporters choose to publish, not the true share of disruptions with a known cause.

Adversary-driven patterns

Six of the twelve COLDGUARD issue classes describe activity that requires an adversary. The other six describe exposure or governance gaps that may or may not have been exploited. Only the first group is counted here.

Adversary-driven findings

59

Published findings in the six adversary-driven issue classes.

Exposure and governance findings

38

Segmentation, monitoring integrity, continuity, configuration, physical dependency, other.

Attributed within adversary-driven

38

High or moderate attribution confidence.

Findings by adversary pattern

n = 97 published findings; dataset demo-2026.11.07. One issue class per finding; pattern labels are editorial names for issue classes.
Categoryfindings
Exploitable software weakness31
Remote access exploitation7
Ransomware and system disruption6
Data theft and exposure6
Credential and identity abuse5
Supplier compromise4
n = 97 published findings; dataset demo-2026.11.07. One issue class per finding; pattern labels are editorial names for issue classes.

Documented impact of adversary-driven findings

n = 59 adversary-driven findings. Unknown means the public material does not state an impact.
Categoryfindings
Unknown29
No documented impact6
Temperature assurance6
Operational interruption5
Inventory integrity5
Clinical service continuity3
Data confidentiality3
Multiple2
n = 59 adversary-driven findings. Unknown means the public material does not state an impact.
Adversary patterns with their underlying issue class, published count, and attributed count
PatternWhat it coversIssue classPublishedAttributedExplore
Exploitable software weaknessA public vulnerability disclosure affecting equipment or software used in the sector.Software vulnerability disclosure3111Findings
Remote access exploitationVendor or staff remote access paths used as an entry point.Remote access governance77Findings
Ransomware and system disruptionExtortion or sabotage that halts monitoring, inventory, or release systems.Ransomware or system disruption65Findings
Data theft and exposureDonor, patient, or customer records taken or exposed.Data protection and privacy66Findings
Credential and identity abuseCompromised, shared, or stale credentials used to reach internal systems.Identity and access55Findings
Supplier compromiseA third party software, logistics, or service provider is the documented origin.Vendor and supply chain44Findings

Exposure and governance classes not counted above: OT / IT segmentation; Monitoring integrity; Incident response and continuity; Configuration or policy disclosure; Physical / digital dependency; Other documented.

Adversary-driven reports by first publication year

Records with a publication date only; dataset demo-2026.11.07. Counts reflect publication, not occurrence.
YearFindings
20194
20205
20218
20227
20236
202411
20255
202613
Records with a publication date only; dataset demo-2026.11.07. Counts reflect publication, not occurrence.

Attribution by sector

Published findings per sector and how many carry high or moderate attribution
SectorPublishedAttributedShare
Blood Banks373184%
Blood Storage191368%
Cold Storage161063%
Food Storage252288%

Sector differences are at least partly reporting differences. See methodology §9 before comparing.

Reference frameworks

Pattern names on this page follow public taxonomies. These are vocabularies, not data feeds: COLDGUARD does not ingest records from them, and nothing above is attributed to them.

What this page will not do

  • Name a group, campaign, or individual, even when a cited source does. Readers can follow the source link on each finding.
  • Infer an attacker from a pattern. A ransomware finding with no stated actor stays unattributed.
  • Publish techniques, tooling, or targeting detail that would help reproduce an attack.
  • Rank countries or organizations by attacker interest. Coverage differences would dominate any such ranking.

Corrections to any attribution level go through the corrections process. The underlying counts are available as JSON at /api/v1/attackers/summary/index.json.