Deepvue Shield

Stop fraud before it enters your system.

Pre‑onboarding risk intelligence: detect loan stacking, block device farms & emulators, score application metadata, and enrich with OSINT – all in milliseconds.

40%

Lower fraud losses

<200ms

Typical API latency

ISO/SOC2

Compliance‑ready

LoanStack

Detect multi‑lender loan stacking & high‑velocity shoppers.

DeviceID

Persistent device fingerprinting; emulator/root detection.

MetaRisk

Behaviour + form‑fill scoring: copy‑paste, speed, referrers.

SignalOSINT

Lead‑portal exposure & reputation signals without a bureau.

What Shield Solves

Loan stacking

Multiple loans in hours

Catch cross‑application bursts, marketplace shoppers, and broker shotgun patterns.

Application risk

Behavioural anomalies

Copy‑paste spikes, ultra‑fast fill, suspicious referrers, domain age risk.

Non‑bureau gaps

Lead‑exposure & OSINT

Signals from phones/emails across lead portals – no competitor data sharing.

Synthetic ID

Device farms & identity reuse

Link devices → identities, spot emulators/VPNs, and stop repeaters.

Product Modules

Start with one module or deploy the full Shield stack. All APIs are stateless, low‑latency, and return reason codes for audits.

Module

LoanStack

Detect multi‑lender loan stacking without becoming a bureau.

Module

DeviceID

Persistent device fingerprinting for web & mobile with emulator/root detection.

Module

MetaRisk

Behavioural & application metadata scoring with explainable reason codes.

Module

SignalOSINT

Non‑bureau enrichments: lead‑marketplace sightings & digital reputation.

Module

MuleCheck

Catch payout/repayment accounts linked to mule activity before disbursement.

See How Shield Works

Follow a borrower from application to decision. Swap in real screenshots later - the structure stays.

Edit Content

Step 1

Borrower applies online

Shield captures device, IP/ASN, referrer/UTM and behaviour metrics on submit.

[ Placeholder: Graph cluster view ]

Edit Content

Step 2

Shield API evaluates risk

LoanStack, DeviceID, MetaRisk and SignalOSINT score the session in <200ms p95.

[ Placeholder: Graph cluster view ]

Edit Content

Step 3

Decision with explanations

Return risk_score, tier and actionable reason codes for audit and CX.

[ Placeholder: Graph cluster view ]

Edit Content

Step 4

Outcome & next steps

Clean borrowers sail through; risky ones go to step‑up KYC or investigation queues.

[ Placeholder: Graph cluster view ]

Simple, fast API

Call Shield synchronously at application submit. Typical p95 < 200ms. Every response includes a score, tier, and explanation.

POST /risk/shield/check
{
  "phone_hash": "sha256(+91xxxxxxxxxx)",
  "email_hash": "sha256(u@example.com)",
  "device_id": "fp_abc123",
  "ip": "1.2.3.4",
  "referrer": "https://loan-compare.example/campaign",
  "utm": {"source":"aff_x","medium":"cpc"},
  "session": {"ttf_ms": 84000, "paste_rate": 0.62, "revisits": 19}
}

200 OK
{
  "risk_score": 0.87,
  "tier": "high",
  "reasons": [
    "Device used for 6 identities in 36h",
    "Recent SIM swap (12h)",
    "Referrer=loan marketplace",
    "High-risk ASN"
  ],
  "actions": ["HOLD", "STEP_UP_KYC", "SELFIE_LIVENESS"],
  "model_version": "shield_v1.3.2"
}

Why teams choose Shield

Starter policy

Recommended thresholds

SIM swap (≤14d) → +40

Device reuse >4/48h → +35

Marketplace referrer → +30

High‑risk IP/ASN → +25

Burst velocity (>3/24h) → +20

New/disposable domain → +15

Calibrate to your hold budget; typical tiers: ≥70 Hold, 50–69 Step‑Up, <50 Pass.

Expected Outcomes

–40%

Fraud loss reduction

+25–35%

Incremental catch vs. rules‑only

<1.5%

Hold rate at scale (typical)

Privacy & Compliance

ISO 27001 Certified

DPDP Ready

Deployment

FAQs

How is Shield different from bureaus?

Shield focuses on non‑bureau signals: device, behaviour, IP/ASN, and OSINT. Optionally, a privacy‑preserving exchange (PSI/Bloom) provides cross‑lender velocity counts without sharing raw PII.

Latency at scale?

API is stateless and horizontally scalable. Typical p95 < 200ms with enrichment caching and adaptive timeouts for third‑party lookups.

Where do AI/LLMs/agents fit?

LLMs generate compliance‑ready narratives; agents run OSINT lookups; ML/GNN powers velocity/device graphs. All outputs include reason codes.

Can we start small?

Yes. Many lenders begin with DeviceID + MetaRisk, then add LoanStack and SignalOSINT as volumes grow.

See Shield in action.

We’ll tailor a walkthrough to your risk stack and portfolio.