A BEHAVIORAL FRAUD LAYER FOR THE CRYPTO ON-RAMP
CRYPTO ON-RAMP SECURITY

We look at the person,
not just the destination.

Address screening catches dirty wallets. It's blind to the most expensive fraud there is — a clean address, and a person being coached in real time to send to it. Straja reads the checkout session itself.

Or read how it works first.

POST /v1/verdictexample session
idle_gaps_over_60s     4
tab_switches           3
address_entry_method   pasted
deposit_escalation     [200, 1000, 4800]
destination_first_seen 2d

INTERVENE  — coached-session pattern, four signals present
75/75 stability runs, zero verdict flips
15 behavioral scenarios calibrated
24h hold at most, never a hard block
HOW IT WORKS

Same checkout. Two completely different sessions.

Address-based screening asks where the money is going. Straja asks what's happening to the person sending it.

ADDRESS SCREENING
“Where is this money going?”

Blocklists, sanctions checks, laundering history. Commoditized — every processor already has this layer.

Blind here: the scammer's address is freshly created and perfectly clean.

STRAJA
“Is this person being coached right now?”

60-second idle gaps, tab switches mid-checkout, a pasted address, deposits stepping from 200 to 4,800 — the shape of someone following instructions from a live call.

No address blocklist can see any of it.

idle_gaps_over_60s
Pauses long enough to suggest someone is being coached off-screen during checkout.
SESSION
address_entry_method
Pasted, typed, or scanned — a pasted address someone else supplied reads differently than a self-scanned wallet.
SESSION
deposit_escalation
The sequence of this account's deposits. A small successful withdrawal, then a large one, is the classic bait-and-scale pattern.
HISTORY
destination_first_seen_days
How new the receiving address is to this specific relationship, not to the wider chain.
DESTINATION

Fifteen signals in all — see the full set.

EVIDENCE, NOT ASSERTIONS

Every number below is measured, not modeled.

The stability suite runs live model calls against fifteen labeled scenarios, five times each, and checks the verdict never moves.

75/75
stability runs, zero verdict flips — including two look-alike-buyer scenarios built to trip a false positive.
$0.044
weighted average cost per scanned transaction, computed from live model pricing.
~2%
of checkout volume ever needs the full-intervention path.
24h
hold on a flagged transfer — never a block. The person can always proceed.
GET STARTED

Run it in shadow mode before you trust it in production.

Four weeks, one payment corridor. Straja scores every session but never intervenes — then you compare against real fraud outcomes before anything goes live.