Everyone scrapes the darknet. We keep it clean.
Adversarial collection makes attribution accurate at the source; crypto-native cleaning makes sure no noise reaches you downstream. Raw darknet collection is mostly noise — donation addresses, decoy "key for sale" scams, decorative wallets, and wallets already labeled in our 450M-address graph. We remove all of it, so what you license is the real money flow.
- Raw scrape — noise shipped as collected
- Donation & decorative addresses inflate the data
- Fake "key for sale" scams ingested as criminal
- Known wallets (e.g. public treasuries) mislabeled
- No site-type classification
- No clustering across sites
- Crypto-native cleaning — high-signal addresses only
- Donation & decorative noise removed
- Scam & exit-scam wallets excluded
- Every address cross-checked against our 450M+ address graph
- Site type classified — non-actionable sites dropped at the source
- Onion sites clustered by shared payment infrastructure
// here's what that looks like on a single scrape
Illustrative sample. We drop donation and decorative addresses, "key for sale" scams and exit-scams, and addresses that are already-known entities (public treasuries, exchange wallets). What remains is high-signal — often flagged before first funding.
How we keep the signal clean
Site-type classification
Our model identifies what a site actually is and drops non-actionable types before an address ever enters the dataset.
Known-entity exclusion
Every address is checked against our 450M+ address attribution graph — a scam "key for sale" that's really a public treasury or exchange wallet is caught and removed.
Donation & decorative noise removed
Surface-level donation and decorative addresses that aren't the real money flow are discounted, not counted.
Scam & exit-scam filtering
Fake key-sale pages and exit-scam wallets are excluded, so the feed reflects real operators — not the fraud around them.
Payment-infrastructure clustering
We cluster onion sites by shared payment rails and processors — linking operations that other providers see as unrelated.
See the difference on your own data.
Request a data license and sample datasets to evaluate DarkScout's clean-signal filtering across the entity types that matter to you.
Request Data License