Stop QR Fraud
Before It Strikes.
AI-powered QR fraud protection trained on Nepal's payment rails. Detects COLLECT-vs-SEND scams, typosquats, and phishing in under 100ms — in Nepali and English.
“You scan to pay — you don't scan to receive.”
तपाईंले पैसा तिर्न मात्र QR स्क्यान गर्नुहोस् — पैसा प्राप्त गर्न होइन।
Active QR Fraud Vectors in Nepal
Documented by Nepal Police Cyber Bureau, NRB, FIU-Nepal. Each is countered by a specific AI defense layer.
COLLECT-vs-SEND Confusion
Fraudsters pose as buyers on Hamrobazar/OLX/Facebook and send sellers a QR 'to confirm payment.' The QR is actually a COLLECT REQUEST — it debits the seller. Nepal's #1 QR fraud type.
Wallet deeplink parsers detect intent at decode time, before the payment screen opens.
Fake NEA / NTC Bill QR
AI-generated fake utility bills with malicious QR codes via WhatsApp / Viber. NPR 45M+ stolen in Kathmandu Valley in 2024 alone.
Document authenticity scoring + institutional QR certificate validation.
Merchant Sticker Overlay
Fake QR stickers placed over real Fonepay/eSewa merchant codes at Thamel shops, restaurants, petrol pumps. Customers pay the fraudster believing they are paying the merchant.
Visual tamper detection compares QR against the merchant's signed certificate.
Daraz / eSewa Phishing QR
Fraudsters impersonating customer service send QR codes to 'verify your account' or 'claim refund.' QR redirects to credential-harvesting login pages.
URL reputation + typosquat detection (Levenshtein vs known wallet hosts).
Dynamic Redirect QR
QR codes using shorteners that change destination after initial inspection — legitimate at scan time, malicious hours later.
Continuous redirect chain monitoring + destination drift detection.
Cross-Border Exploitation
Indian UPI payments at 1.5 lakh+ Fonepay merchants since 2024; 10-country NEPALPAY tourist QR. Fraudsters exploit the weaker verification on cross-border rails.
Joint fraud signal sharing with NPCI and Alipay (post-funding).
How QRKavach works
Five stages, sub-100ms end-to-end. Heuristic layer always runs; ML layer adds depth.
Capture
User scans a QR via camera, upload, or screenshot. Pixels intercepted before any URL opens.
Decode
EMV-TLV parser handles Fonepay/NEPALPAY merchant codes. Wallet deeplink parsers (eSewa, Khalti, IME Pay) extract intent — SEND vs COLLECT.
AI Analysis
Multi-layer model scores the payload: URL reputation, domain age, typosquat distance, redirect chain drift, Nepali phishing NLP.
Threat Intel
Cross-referenced against VirusTotal, PhishTank, APWG, and Cyber Bureau incident feeds (post-funding).
Decision
Risk score → Safe / Warn / Danger. Verdict surfaced in Nepali + English with the exact reason.
Nepal's payment rails need a security layer.
We're building it.
Pre-funding round open. Looking for investors and NRB / fintech partners ready to back AI fraud protection for the world's fastest-growing QR market.