Hudson Data · Centurion + HD Trust
Real-time credit and fraud decisions, in one system.
Hudson Data helps regulated teams approve more good customers, stop more fraud, and protect legitimate users when the session itself is under attack.
Real-time decisioning · SOC 2 Type 2 · Forward-deployed delivery
Live demo · same session, two views
Privacy-conscious customer · Brave + Tor
Fingerprinting platform
Sees a suspicious browser.
- Device✗ Tor exit
- Browser⚠ Anti-fingerprint
- IP✗ Tor exit node
- Fraud score✗ 820 / 1000
Verdict
BLOCK
Your loyal customer just got locked out for using privacy tools.
HD Trust
Sees a known customer.
- WHO✓ LEGIT_OWNER
- WHAT⚠ PRIVACY
- HOW⚠ TOR_EXIT
- Action✓ APPROVE
Verdict
APPROVE
Privacy browser, but account behavior confirms the owner.
Fingerprinting blocks your customer. HD Trust knows them.
The three axes
WHO × WHAT × HOW in one decision.
HD Trust resolves the operator, the device and runtime, and the connection separately, then turns that structure into the action your policy calls for.
WHO
Operator identity + behavioral signal
Who is operating the session — and how are they acting?
Identity continuity, behavior, and history determine whether the operator is the account owner, a bot, a mule, or a customer under pressure.
Resolves to
LEGIT_OWNERCOACHEDSTOLEN_IDMULEBOT_SCRIPTAI_AGENTWHAT
Device, browser, runtime integrity
What is the device — and is it authentic?
Runtime and browser integrity show whether the environment is authentic, privacy-conscious, compromised, automated, or built for stealth.
Resolves to
AUTHENTICPRIVACYCOMPROMISEDHEADLESSSTEALTHAI_BROWSERHOW
Connection integrity + transport
How is the session reaching you — and is it intact?
Transport and network signal show whether the connection is direct, privacy-routed, proxied, or being tampered with.
Resolves to
DIRECTPRIVACY_VPNPROXYTAMPEREDHow the verdict is decided
Decide from structure, not a single score.
A single fraud score hides too much. HD Trust returns a WHO × WHAT × HOW matrix so teams can distinguish a bad actor from a good customer in a bad environment. The verdict comes from the matrix cell, with policy controls by organization and action.
| WHO ↓ / WHAT → | AUTHENTIC | PRIVACY | COMPROMISED | HEADLESS | STEALTH | AI BROWSER |
|---|---|---|---|---|---|---|
| LEGIT. OWNER | APPROVE | APPROVE | PROTECT | STEP-UP | STEP-UP | REVIEW |
| COACHED | REVIEW | REVIEW | BLOCK | BLOCK | BLOCK | BLOCK |
| STOLEN ID | STEP-UP | STEP-UP | BLOCK | BLOCK | BLOCK | BLOCK |
| MULE | REVIEW | REVIEW | BLOCK | BLOCK | BLOCK | BLOCK |
| BOT SCRIPT | BLOCK | BLOCK | BLOCK | BLOCK | BLOCK | BLOCK |
| AI AGENT | REVIEW | REVIEW | BLOCK | BLOCK | BLOCK | REVIEW |
Why PROTECT matters
When the operator is legitimate but the device is compromised, the right answer is not always a block. PROTECT warns the customer, steps up authentication, and preserves the relationship while reducing loss.
Why the matrix helps
A legitimate owner on a compromised device and a stolen identity on a clean device can look similar in a device-only workflow. The matrix separates those cases so operations, fraud, and customer-care teams can act differently.
The matrix above is an example default. Each cell is tunable by product, channel, action, and risk appetite.
Session-level threat coverage
Session-level threats need session-level signal.
Modern fraud blends automation, coercion, malware, proxy infrastructure, and real customer behavior. HD Trust is designed to preserve that context through the decision.
AI-agent browser attack
Automation controlling a real browser to apply for credit or transact.
Voice-coercion fraud
A customer being coached through a high-risk action in real time.
Remote-access scam
Remote-control software operating on the customer's own device.
Push-payment scam
A legitimate user is pressured into sending funds to a fraudulent destination.
Stealth browser
A coherent browser profile built to evade device-only checks.
Malware-mediated theft
Browser overlays, injected fields, or background actions inside a trusted session.
MITM / proxy injection
Traffic interception, replay, or tampering between client and service.
Synthetic-identity ring
Coordinated profiles and devices assembling fake applications at scale.
In production at scale
$2.5B+
annual transactions scored
Real-time
session decisioning
SOC 2 Type 2
certified
When you need more than session trust
Centurion is the broader decisioning fabric.
HD Trust is the session-trust layer. Centurion Platform connects credit models, network-fraud signal, and production decisioning so teams can adopt one layer or the full system.
Pillar 01 — Credit
Build credit models. Run them in production.
Model Foundry builds underwriting, behavioral, and operational risk models from pre-built recipes. FlowX runs them in production with versioning, monitoring, and instant rollback.
- 1Model Foundry — governed credit-model development factory
- 2FlowX — production decisioning, also where fraud signals land
Pillar 02 — Fraud
Detect fraud in real time. At the moment of decision.
ML Graph scores an entity's place in the fraud network. HD Trust scores the live session. Both feed FlowX so fraud actions and credit decisions share the same policy layer.
- 1ML Graph — real-time fraud graph at scale
- 2HD Trust — session trust intelligence, three-axis scoring
Holistic underwriting
One system. Credit and fraud, end to end.
FlowX is where the credit and fraud signals converge. Approve when both signals are clean. Protect the legitimate user on a compromised device. Step up when uncertain. Block the adversary. One decision engine — informed by every signal you have.
01 · Customer event
Trigger
- Loan application
- Transaction
- Account opening
- Login & payout
- Account change
02 · Centurion scores
Credit + fraud · sub-second
Fraud
ML Graph
Real-time fraud graph
HD Trust
Session trust signal
Credit
Model Foundry
Underwriting & behavioral models
FlowX
Unified decisioning
versioned · roll-back · audit
03 · Decision
With reason + audit trail
Legitimate user · clean device
Legitimate user · compromised device — warn + step up auth, don't block
Uncertain — verify identity before proceeding
Adversary — reject and log as fraud
Every decision returns with feature snapshot, strategy version, and rationale — replayable years later.
Your data
Loaded once · streams in real time
External signals
Bureau / vendor APIs · wired into FlowX
One platform that combines credit and fraud signals into a single decision — so you approve more good business and block more bad business at the same time.
Professional Services
A senior team behind every deployment.
Centurion ships with senior practitioners who configure the models, policies, integrations, and governance needed to move from promising signal to production decisions.
Custom credit models
Underwriting, behavioral scoring, collections, capital — built on your portfolio with our recipe library and senior modelers.
Fraud rules & detection logic
Custom detection strategies for your fraud surfaces: account opening, transaction, ATO, mule, synthetic identity.
Integration engineering
Bureau, KYC, identity, payments, sanctions vendors — wired into FlowX with retries, budgets, and graceful degradation.
Risk advisory
When labels are delayed, signal is noisy, or policy scrutiny is high, our team helps turn ambiguity into defensible operating decisions.
Use cases
Risk decisioning, across the customer journey.
Credit decisions span every lending product. Fraud decisions live everywhere money or value moves. Centurion deploys at every point in your stack where one of those decisions is made.
01
Attract
Prospecting
02
Acquire
Apply & onboard
03
Engage
Authenticate & transact
04
Grow
Service & expand
05
Resolve
Collect & recover
Selected work
In production.
01 /
Fintech
Credit + fraud
production decisioning
Production underwriting and fraud models for portfolios where growth, loss control, and explainability all matter.
02 /
Insurance
Policy + claims
fraud detection
Claims-fraud detection that surfaces coordinated patterns across claims, devices, providers, and policy history.
03 /
Banking
Graph + session
real-time controls
Real-time graph detection that tracks tainted assets and suspicious transaction paths before losses scale.
Why Hudson Data
Built for adversarial, regulated environments.
Credit and fraud, unified
Most stacks treat credit risk and fraud as separate decisions. We treat them as one — better approvals, fewer losses, fewer false positives at the same time.
Real-time graph at scale
ML Graph scores entities at the moment of decision — caught at transaction time, not in next quarter's case-review queue.
Domain depth, not generic ML
Twenty years of risk and fraud across credit, insurance, and banking. The recipes, the heuristics, and the failure modes are baked in.
Cybersecurity in the DNA
Bot detection, ATO defense, stealth-browser fingerprinting, behavioral analysis. The trust layer is built by people who've spent careers on adversarial systems.
SOC 2 Type 2 certified
Audited controls across security, availability, and confidentiality. We sign MSA, BAA, DPA, and vendor reviews as needed — engagement-dependent.
Decision excellence as the KPI
Headline AUC isn't the goal — fewer bad decisions in production is. Every system ships with the eval harness, drift monitoring, and decision-quality reporting to prove it.
Start the conversation
Let's talk through your decisioning problem.
A 30-minute conversation usually surfaces whether Centurion is the right fit and where it'd land first in your stack.