methodology · text last changed on 14 November 2026

How the counts are made

Every number on the data pages is a count of collected records. This page says what a record is, how its fields are read, and what the counts cannot tell you.

What a record is

A record is one post that AdverseMonitor collected from a source: a claim, a listing or an announcement about a cyber attack. It is stored with the fields the collection gives it: a category, the actor names, the countries and the industries it names, the type of source, and the time it was collected.

A record is a lead to review, not a confirmed incident. A post can be wrong, exaggerated or repeated, and AdverseMonitor has not checked the claims behind these counts.

The data pages never show a single record. They do not name an organisation, quote a title, show a site or link to a source.

The three source types

A record comes from one of three types of source: Telegram, the open web, or Tor. The type is stored for 33% of the records (23,747 Telegram, 20,395 open web, 10,030 Tor). Older records were stored without it.

Sources can disappear, restrict access or change format. The collection does not cover every source, and AdverseMonitor does not publish how often each one is read.

How the fields are read

The pages leave the stored values as they are and clean them while reading, by fixed rules in code:

  • Category. Every record has one. Ten are named: DDoS Attack, Data Breach, Ransomware, Defacement, Initial Access, Data Leak, Alert, Malware, Vulnerability, Cyber Attack. A stored category that is none of these counts as Other.
  • Countries. A record can name none, one or several. 30% of the records name at least one. "USA" and "United States" are one country, as are "UK" and "United Kingdom". A country page counts the records that name the country. It says nothing about who is responsible for anything.
  • Industries. A record can name several. 27% name at least one. The industry names are the stored ones, so some overlap, for example Government Administration and Government & Public Sector.
  • Actor names. 33% of the records name an actor. The name is the one the source gives. Two spellings are counted as one name only when they differ in upper and lower case, spacing or plain separators. Names that differ in any other way stay apart. The pages do not say who is behind a name, where it is from or what it wants.
  • Time. A record is counted in the month it was collected, in UTC. That can be later than the day the post was published.

The period covered

The data set starts in March 2023, the first month with at least 100 records. 198 records with an earlier collection date are left out, because those months hold too few records to describe. It holds 162,751 records up to 21 November 2026.

The collection has grown over time. More records in a later month can mean more posts. It can also mean that the collection read more sources.

What the counts do not mean

  • A count is not a number of attacks or incidents. It is a number of collected posts.
  • The same claim can be posted in more than one place and then counts more than once.
  • A count of zero does not mean nothing happened. It means no collected record names it.
  • A name that appears often is a name that is posted often in the sources that were read.
  • "Similar counts" on an actor page compares numbers. It does not say two actors are connected.

Which names get a page

A name gets a page when at least 5 records name it. Search engines are asked to index the page only when it has enough data to be worth reading: 30 records for an actor, 50 for a country or an industry. Below that the page exists but is marked as not for indexing. It changes by itself when the count grows.

How often the pages update

Code recounts all pages about once an hour. This count is from 21 November 2026. The last full count read 162,949 records in 6.0 seconds.

A monthly report appears on the first day after the month ends and does not change after that. The yearly report grows with each finished month.

Who writes the text

Code puts the sentences on the data pages together from the numbers of each page. No AI model is involved. If a page ever carries a written summary, it stands in a marked box that says who wrote it and when.

How to ask for a correction

If a count, a name or a spelling on these pages is wrong, write to us through the contact page. Give the address of the page and what is wrong. The pages hold no record details, so a correction is about how something is counted or named.

More about how the product matches records to a company: the product methodology.

Back to the data overview or the threat actors.