Threat level: elevated · 14,302 transactions blocked today
Fraud doesn't
announce itself.
AURA doesn't either.
ML catches what rules miss. Rules catch what ML can't explain. Every transaction scored in under 80ms — automatically actioned, fully auditable, yours to tune.
Fraud blocked · 30d
$12.4M
↑ threats rising
Catch rate
99.3%
industry avg 94.1%
False positive rate
0.08%
10× below baseline
Avg decision time
67ms
p99 · all signals
Threat intelligence
Live threat surface. Global visibility.
Every flagged transaction mapped to origin. Every blocked attempt logged with full signal context.
Global threat heatmap
Last 60 min
Threat intensity:
None
Elevated
High
Live decision feed
live
0.91
txn_4829_card · $4,210
Signals: geo+vel+bin
Block
now
0.12
txn_4830_sepa · $890
Signals: clean
Clear
2s
0.67
txn_4831_wire · $12,400
Signals: vel+device
Review
5s
0.04
txn_4832_card · $220
Signals: clean
Clear
8s
0.88
txn_4833_ach · $7,800
Signals: geo+aml
Block
11s
0.51
txn_4834_card · $3,100
Signals: time+bin
Review
14s
Signal engine
Eight signals. One decision. 67ms.
Every transaction is decomposed into weighted signals. The model explains every score — no black boxes, full audit trail.
Explainable by design.
CTOs and risk teams don't just get a score — they get a ranked breakdown of every contributing signal, with confidence intervals and the ability to override any weight.
Every decision stored with full signal snapshot
Model retrained weekly on your transaction corpus
Override any signal weight from the dashboard
Exportable to your data warehouse via API
Transaction · txn_4829_card
0.91
HIGH RISK
Decision: BLOCK · 67ms
Amount
$4,210
Card BIN · RU
Geo mismatch
0.94
HIGH
Velocity 1h
0.87
HIGH
Device fingerprint
0.72
MED
BIN risk score
0.81
HIGH
Behavioural pattern
0.44
MED
AML watchlist
0.12
LOW
Network graph
0.31
LOW
Time anomaly
0.19
LOW
Rule engine
Your risk. Your rules.
ML catches the novel threats. Your rules handle the ones you already know about. Build, test, and deploy logic in minutes — no code required.
Full control, zero engineering dependency.
Define thresholds, combine signals, set actions. Rules run in parallel with the
ML model — and you decide which takes priority when they conflict.
✓
If/then logic across 40+ transaction attributes
✓
Dry-run mode: test against 30-day history before deploying
✓
Rule versioning with instant rollback
✓
Conflict resolution: rule-first or model-first per ruleset
✓
Alert threshold and escalation routing built in
Rule editor
dry-run active
v2.4
IF all conditions match
01
velocity_1h
≥
5
transactions
02
geo_mismatch
=
true
03
amount_usd
>
2000
Then
Block
Review
Step-up
Dry-run: 342 matches in last 30d · 0 false positives
Deploy
Response actions
Four outcomes. Zero ambiguity.
Every flagged transaction gets an immediate, deterministic action. No queue, no delay, no manual triage required.
Block
< 5ms · hard decline
Transaction terminated at the edge. Card issuer receives a decline code. Customer sees a generic error. No partial auth, no retry window.
Manual review
< 200ms · queue + notify
Payment held. Risk analyst notified via webhook and dashboard. SLA timer starts. Auto-releases to approve or block after configurable window.
Step-up auth
< 80ms · 3DS2 challenge
Customer prompted for additional verification — biometric, OTP, or 3DS2 challenge. Passes without friction if successful. Flagged if it fails.
Allow + monitor
< 10ms · shadow mode
Transaction clears but enters enhanced monitoring. Subsequent behavior patterns logged and fed back to the ML model for retraining.
Performance
Numbers a CTO actually cares about.
Measured against your transaction corpus. Reported in your dashboard. Contractually backed at Enterprise tier.
99.3
%
Fraud catch rate · 30d trailing
Verified against confirmed fraud reports. Includes card-not-present, account takeover, and synthetic identity vectors.
0.08
%
False positive rate · 30d trailing
Legitimate transactions incorrectly flagged. Industry baseline is 0.6–1.2%. Every false positive is reviewed and fed back into the model.
67
ms
p99 decision latency · global
Measured end-to-end from transaction receipt to action dispatch. All 8 signals evaluated in parallel, not sequence.
See every signal. Control every decision.
Full fraud engine. Live in your stack today.
No credit card required · Sandbox access instant · Model tuned to your data in 7 days