Real-time fraud intelligence
Every transaction.
Scored in 47ms
Before money moves.

Flint scores every transaction in under 50ms using device intelligence and behavioral analysis. One API call, before money moves.

View Demo →
status: live

OTP is not fraud detection.

Digital payments in South Asia are growing 40-50% year over year. Fraud is growing faster. SIM swaps, social engineering, money mule networks, authorized push payment fraud — attacks that OTP was never designed to stop.

OTP answers one question: do you have this phone number? It says nothing about the device, the behavior, the timing, the recipient, or the pattern. When an attacker swaps your SIM, OTP helps them. When a victim is socially engineered into reading their OTP aloud, OTP is irrelevant.

Most payment companies rely on blanket rules — block everything above a threshold, require OTP for every transaction, freeze accounts after failed logins. These rules block legitimate users as often as they block fraudsters.

Rs 50L+

Lost to fraud annually by a single mid-size PSP

73%

of fraud bypasses OTP-only protection

5-10x

higher false positive rates with rule-based systems

Flint replaces blanket rules with intelligent, per-transaction decisions. 90% of users experience zero friction. Fraudsters get blocked before money moves.

One API call, one decision

Your system calls our API before processing any payment. We return a score and decision. If we don't respond in time, your system fails-open to ensure no user is ever blocked by us.

01

User's Device

Flint SDK inside

Collects 50+ device signals

02

Your Server

Calls Flint API

Sends payload before processing

03

Flint Intelligence

Score in <50ms

Runs behavioral & network models

04

Decision

ALLOW / FLAG / BLOCK

Immediate response to your server

ALLOW

Risk Score: 0-50

Transaction proceeds instantly. User notices nothing. This covers 90% of all transactions.

FLAG

Risk Score: 50-80

User sees a lightweight verification step like an OTP. Catches suspicion without hard-blocking.

BLOCK

Risk Score: 80-100

Transaction stopped before money moves. Only triggered when multiple signals agree on high risk.

Score a transaction. See the logic.

Adjust the parameters below and watch how the risk score changes. This is a simplified version of the scoring logic. The production engine uses 50+ signals and per-user behavioral profiles.

Rs 5,000
2:00 PM
1 tx/hr
3/ 100
ALLOW
Device Assessment
0/35
Behavioral Analysis
3/25
Velocity Check
0/20
Recipient Risk
0/20

Low risk: Trusted device attempting Rs 5,000 transfer at 2:00 PM. Transaction proceeds without friction.

How we actually catch fraud

Each transaction is analyzed across device intelligence, behavioral patterns, and cross-platform network data. Combined into a single risk score in under 50ms.

fp_3f8a...9d
trust: 92
seen: 47x

Device Intelligence

50+ signals. One fingerprint.

A lightweight SDK collects device characteristics — screen resolution, GPU rendering patterns, audio processing fingerprint, installed fonts, browser configuration. No biometrics. No personal data. Just observable hardware and software signals that combine into a unique device fingerprint. Devices build trust over time. A new device starts at low trust and earns confidence through days of normal use.
  • Canvas and WebGL rendering fingerprint
  • Audio processing fingerprint
  • Font enumeration
  • Bot and emulator detection
  • Progressive trust scoring
6am6pm12am

Behavioral Analysis

Per-user baselines. Real-time deviation.

Flint builds a behavioral profile for every user — average transaction amount, active hours, common recipients, typical velocity, spending rhythm. When a transaction deviates from the user's established pattern, the risk score increases proportionally. Rs 5,000 at 2 AM is suspicious for someone who averages Rs 1,200 during business hours. Flint adapts to individual behavior, not static rules.
  • 30-day rolling behavioral profiles
  • Amount anomaly detection (z-score based)
  • Active hours histogram per user
  • Velocity baseline comparison
  • Transaction type pattern analysis

Network Intelligence

Caught once. Blocked everywhere.

When a device or account is confirmed as fraudulent on any client's platform, that signal propagates across every Flint client in under 300 milliseconds. A fraudster caught on one payment app is immediately blocked on every other payment app, bank, and e-commerce platform in the network. Intelligence sharing ensures everyone benefits from catching a fraudster once.
  • Cross-client device blacklisting
  • Real-time fraud signal propagation
  • Money mule chain detection
  • Recipient risk network analysis
  • Anonymized intelligence sharing

Drop it into your stack

Add our SDK to your app, add one API call to your backend. If Flint disappears tomorrow, your system works exactly as before. Device fingerprinting SDKs are open source under the MIT license — read every line on GitHub.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
// 1. Add SDK to your app (one time)
// <script src="https://cdn.flintsecure.app/v1/fp.js"></script>

// 2. Get device fingerprint
const fp = await FlintFP.load({ apiKey: 'YOUR_KEY' })
const { fingerprint } = await fp.get()

// 3. Score transaction before processing
const response = await fetch('https://api.flintsecure.app/v1/score', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    sender_id: user.id,
    amount: 5000,
    device_fingerprint: fingerprint,
    receiver_id: recipient.id
  })
})

const { decision } = await response.json()

// 4. Act on the decision
if (decision === 'ALLOW') processPayment()
if (decision === 'FLAG')  requestOTP()
if (decision === 'BLOCK') rejectTransaction()

Fail-Open by Design

If Flint is slow or down, your transactions proceed normally. We set a 150ms timeout. If we don't respond, your system continues as if Flint doesn't exist. Flint can never cause a payment to fail.

Zero Infrastructure Change

No database changes. No code rewrites. No infrastructure to manage. Add a few lines of code to your app and backend. Everything runs on Flint's infrastructure. Remove it anytime.

Where every millisecond goes

Speed is a security feature. We optimized our stack to ensure fraud detection never slows down your checkout flow.

Network
15ms
Validate
0.1ms
Redis
2ms
Score
0.5ms
Respond
0.1ms

Total processing time: 2.7ms

Network overhead: 15ms

<50ms

Median Response

P99 under 100ms

0ms

Failure Rate

Fail-open architecture

6

Parallel Lookups

Redis in-memory

0

DB Queries

RAM-first architecture

Pricing for every scale

Every plan includes the full scoring API and device intelligence. Unlock network intelligence and behavioral profiling as you grow.

Developer

Freeforever

For testing and evaluation.

  • 1,000 API calls / month
  • Device fingerprinting
  • Basic risk scoring
  • Community support
  • Network intelligence
  • Behavioral profiling
  • Dashboard
MOST POPULAR

Growth

Rs 1,00,000/ month

For scaling payment apps.

  • 500,000 API calls / month
  • Full scoring engine
  • Network intelligence
  • Behavioral profiling
  • Full dashboard
  • Webhook alerts
  • Custom thresholds
  • Priority support

Enterprise

Customnegotiated

For high-volume platforms.

  • Unlimited API calls
  • Everything in Growth
  • Custom ML model training
  • Dedicated account manager
  • SLA: 99.99% uptime
  • On-premise deployment
  • Custom integrations
  • 24/7 Phone support

Shadow Mode TrialAll plans include a 3-month trial in shadow mode. Flint scores every transaction but blocks nothing. You review the numbers. If the detection accuracy doesn't convince you, remove Flint at zero cost.