Skip to content

Guides

Track ransomware leak sites programmatically

When a ransomware group claims a victim, the first public evidence is usually a post on the group's leak site - often days before any press coverage or disclosure. If you do vendor risk, insurance, journalism or sector monitoring, that stream is checkable programmatically.

What leak-site data looks like

Each listing is roughly: group, victim name, date discovered, country, sector, and whether it was later delisted. Caveats that matter and that most people skip:

  1. A listing is a claim, not a confirmed breach.
  2. Groups repost, exaggerate and sometimes list the same victim twice under rebrands.
  3. Delisting is signal, not proof of payment.

Querying it live

ThreatCluster monitors the leak sites continuously (Tor crawlers plus validation). The API returns listings filtered by group, sector, country or window - free keys are 100 credits a day (get one here):

# Who did Akira list in the last 30 days?
curl -s "https://threatcluster.io/api/public/v1/darkweb/ransomware/victims?group=akira&days=30" \
  -H "X-API-Key: tc_live_..."

There is also an RSS feed if you just want it in a reader, the browsable tracker at /dark-web, and live aggregate figures - victims by month, group, sector and country - at /dark-web/statistics.

The open dataset (20k+ listings, CC-BY)

For analysis rather than monitoring - victim counts per group over time, sector distributions, inter-listing intervals - there is a CC-BY-4.0 dataset of 20,000+ historical listings on Hugging Face:

from datasets import load_dataset
ds = load_dataset("threatcluster/ransomware-leak-site-victims")

Fields include group, sector, country, timing and per-group victim sequence numbers, plus an is_headline_style flag for the ~0.5% of rows that are post headlines rather than clean organisation names - filter those before publishing anything.

For AI agents

Same data as an MCP tool (leak_site_victims): npx -y threatcluster-mcp, or one click from the Claude directory.

Ethics note

Leak-site data names real victim organisations. The dataset deliberately excludes victim descriptions (extortion copy), post URLs and screenshots; respect that line in anything you build on it.