CSI Atlas

Quality

What is measured here, with its denominator, and what a check does not prove.

The scorecard

This page is generated from the published records each time it is served. Every measure shows its denominator. A high value on one row does not compensate for a low value on another.

Quality scorecard, generated from the published records.
MeasureValueDenominatorWhat it does not tell you
Published findings195records in this publicationCounts published records only. It is not a count of incidents.
Findings that cite a source195 (100%)of 195 published findingsA cited source is a starting point. It does not confirm the claim.
Publishers cited25distinct publishersA publisher appearing often is not independent corroboration.
Findings with a stated event date12 (6%)of 195 published findingsThe rest keep a blank event date rather than an inferred one.
Findings with a retained original189of 195 published findingsA retained original proves the bytes we read, not the truth of the claim.
Event classes with findings8 of 10ten classes in the taxonomyA class with no findings is a coverage gap, not evidence of absence.
Classes with 30 or more findings2 of 10the chart floor is 30Below the floor a class shows its count instead of a chart.
Findings awaiting a class2of 195 published findingsPublic intelligence records wait for a classifier decision rather than a guess.
Corrections in this snapshot0entries in the public logNo entry means no correction was issued. It does not mean no error exists.

How to read it

These are checks on process, not on the world. They tell you whether a record kept its source, whether its date was stated, and whether the class could be assigned. They do not tell you whether a reported loss is accurate, whether an actor is responsible, or whether a vulnerability was actually exploited.

Where a measure is low, the site says so rather than hiding it. The largest known gap is stated on the same page, with the count behind it.

Known gaps in this snapshot

183 of 195 published findings do not state an event date, because the cited source did not state one. Charts therefore place records by their update time, and the event date is shown only when a source stated it. 2 findings have no assigned class and are counted as awaiting a class. Jurisdiction is unknown for 190 findings.

A passing check is not proof

A check can pass and a claim can still be wrong. Source matching confirms wording and date against the retained copy. It does not independently verify an actor, an amount or an attribution. Those remain reported claims until a stronger source supports them, and the uncertainty line on each record says so.