PayeeCheck · Fraud API · India

Stop UPI Fraud
in 2.2ms

India's first AI-powered payee verification. 7 signals. Real-time decisions for the ₹17 billion moving on UPI every month.

UPI has no mandatory name verification. PayeeCheck adds it.
₹17B monthly UPI
·
7 fraud signals
·
2.2ms pipeline
·
100% precision
No account needed Runs in browser Takes 60 seconds
01
🏦
Initiate transfer
Enter payee name, UPI ID, and amount like a real transaction
→
02
⚡
7 signals fire
Name matching, mule DB, VPA velocity, behavioral SDK — all in 2.2ms
→
03
🛑
Fraud blocked
See exactly which signals fired and why the transaction was stopped
See the 7 signals
How it works

7 signals. Every transaction.

Each layer catches what the last one misses. Together they achieve 100% precision on the scenarios that matter most.

🔤
Name Matching (L1–L5)
Jaro-Winkler → Phonetic → TF-IDF → Embeddings → Siamese neural net. Handles transliteration, abbreviations, and entity-type mismatches.
WARN if < 0.85
🎭
Lookalike VPA
Brand token registry catches impersonation — "paytm.support@", "hdfc-help@", "k.enterprises@" — without ML overhead.
BLOCK on match
🕳️
Mule Database
Consortium-style flagging across partner banks. Ring propagation scoring — one flagged entity lifts the whole ring.
BLOCK if flagged
📱
Behavioral SDK
Paste vs. type input method, keystroke cadence, device fingerprint, clipboard timing. Outputs a trust score 0–1.
WARN if paste
⚡
VPA Velocity Engine
Rolling-window anomaly detection — new VPA with high unique sender count is a strong fraud ring signal.
WARN if < 30d
🚫
Sanctions Screening
Fuzzy name matching against PEP and OFAC lists with phonetic normalisation to catch spelling variants.
BLOCK on match
🕸️
GAT Ring Detector
PyTorch Graph Attention Network. Detects fraud rings sharing mobile numbers, accounts, or devices. 6.7pp edge feature lift.
BLOCK on ring
2.2ms
pipeline latency
100%
precision
82.6%
recall
7
signal layers
See all 7 signals fire in real-time · No account needed
The problem

India sends ₹17B daily on UPI.
None of it verifies who's receiving it.

The UK has had mandatory Confirmation of Payee since 2019. The EU mandated Verification of Payee in 2025. India — the world's largest real-time payments market — still has no equivalent. PayeeCheck fills that gap today.

United Kingdom
Confirmation of Payee (CoP)
✓ Live since 2019
  • Mandatory for all major UK banks
  • Real-time account name check before payment
  • APP fraud reduced 46% post-implementation
  • Regulated by Pay.UK under PSR
European Union
Verification of Payee (VoP)
⏳ Mandated from 2025
  • Required under EU Instant Payments Regulation
  • All PSPs must verify payee name before SEPA transfers
  • Covers 36 countries across SEPA zone
  • Non-compliance results in regulatory penalties
India
No mandate. No standard.
✗ No regulation yet
  • UPI processed ₹20L crore monthly in 2024
  • Zero requirement to verify payee name
  • Fraud losses crossing ₹11,000 crore annually
  • RBI advisory guidance only — no enforcement
₹11,000Cr
Annual UPI fraud losses in India (RBI, 2024)
0
Regulatory requirements for payee name verification on UPI
2.2ms
Time PayeeCheck takes to run all 7 signals per transaction
Developer-first

One API call. Seven signals. Production-ready.

Drop PayeeCheck into any payment flow in minutes. POST a transaction context, get a grounded risk verdict with auditable signal assertions — every time.

POST 
curl -X POST https://payeecheck.vercel.app/attribution \
  -H "Content-Type: application/json" \
  -d '{
  "transaction": {
    "input_method":      "paste",
    "paste_trust_score": 0.08,
    "amount":            250000
  },
  "payee": {
    "vpa_age_days":   11,
    "mule_flagged":   true,
    "mule_flag_count": 3,
    "name_score":     0.40
  },
  "temporal": {
    "first_time_payee":    true,
    "amount_vs_avg_ratio": 4.5
  }
'
200 OK /attribution response 2.2ms 
{
  "verdict":              "BLOCK",
  "risk_score":           87,
  "total_signals_fired":  5,
  "grounded_assertions": [
    {
      "signal":    "clipboard_paste",
      "assertion": "VPA pasted — social engineering risk"
    },
    {
      "signal":    "mule_account",
      "assertion": "3 consortium flags on this VPA"
    },
    {
      "signal":    "new_vpa",
      "assertion": "VPA registered 11 days ago"
    },
    {
      "signal":    "high_value",
      "assertion": "₹2.5L is 4.5× above average"
    },
    {
      "signal":    "weak_name_match",
      "assertion": "Score 0.40 — below threshold 0.85"
    }
  ]
}

Built for every player in India's payment stack

Scheduled Banks
Add payee name verification to RTGS, NEFT, and IMPS outward transfers. Meet RBI fraud reporting requirements with grounded audit trails.
Payment Gateways
Screen payouts before settlement. Catch mule accounts and lookalike VPAs in the 200ms window before UPI confirmation is sent.
Neobanks & Fintechs
Ship fraud-aware UPI transfers with a single API call. Behavioral SDK captures paste vs type signals from your existing frontend — no SDK install needed.
NBFCs & Lenders
Validate loan disbursement accounts before payout. Consortium mule database catches fraudulent beneficiaries missed by standard KYC flows.
E-commerce & Marketplaces
Protect seller payouts and refunds. VPA velocity engine flags newly-created accounts receiving high-value payouts for the first time.
Wallet Providers
Layer fraud detection on P2P transfers and merchant payouts. GAT ring detector surfaces connected fraud networks invisible to rule-based engines.
Available endpoints
Full API reference at /docs
POST/compare
POST/match/{level}
POST/attribution
POST/resolve
GET/rings
GET/health
Try the technical sandbox
Run name matchers, entity resolver, and fraud ring scanner live — no auth needed
Open Sandbox
PayeeCheck · Live Demo
0 pts
Step 1 of 5 · Initiate Transfer

Who are you paying?

Start a simulated UPI transfer. Type the payee name — PayeeCheck's Behavioral SDK begins analyzing immediately.

Payee Name
Behavioral SDK
Active
status Waiting for input…
Try:
Step 2 of 5 · Payee Identity

Choose a UPI address

Select or paste the payee's UPI ID. VPA Velocity Engine checks registration age, transaction count, and mule database hits.

🏪
K. Enterprises
k.enterprises@oksbi
⚠ Suspicious
✅
Krishna Enterprises (Official)
krishna.official@ybl
✓ Verified
Or paste / type a UPI ID
VPA Velocity Engine
Active
status Select or enter a UPI ID…
Step 3 of 5 · Transaction Details

How much are you sending?

Enter the amount. PayeeCheck cross-references against your transaction history and captures device fingerprint signals in real-time.

Amount (₹)
₹
Risk Level MEDIUM
Device & Temporal Signals
Capturing
Initialising pipeline…
PayeeCheck Running
7 signals · sequential pipeline
0.0
ms latency
🛑
BLOCKED
PayeeCheck stopped this transaction
₹2,50,000
protected from potential fraud
Signals that triggered
✓
₹2,50,000 protected · Transaction blocked in 2.2ms

The PayeeCheck System

7 fraud signals · Everything that ran in 2.2ms, now in full detail

🔤
Name Matching (L1–L5)
Multi-stage matchers: Jaro-Winkler → Phonetic → TF-IDF → Embeddings → Siamese neural net. Handles Indian transliteration and abbreviations.
🎭
Lookalike VPA
Brand token registry pattern detection. Catches impersonation like "paytm.support@" or "k.enterprises@" without ML overhead.
🕳️
Mule Database
Consortium-style flagging across partner banks. Ring propagation scoring — if one entity in a cluster is flagged, the whole ring scores higher.
📱
Behavioral SDK
Client-side analysis: paste vs. type input method, keystroke cadence, device fingerprint, clipboard timing. Trust score 0–1.
⚡
VPA Velocity Engine
Rolling-window anomaly detection. New VPA with high transaction volume = red flag. Seasonal anomaly flagging built in.
🕸️
GAT Ring Detector
PyTorch-based Graph Attention Network. Detects fraud rings sharing identifiers — mobile, account, device. 6.7pp edge feature lift.
Explore the full technical sandbox
Run the L1–L5 name matchers, entity resolver, and fraud ring scanner — with live results
Open PayeeCheck
2.2ms
pipeline latency
100%
precision
82.6%
recall
7
fraud signals