📦 Fraud Detection SDK

A lightweight JavaScript SDK to interact with a Machine Learning–powered fraud detection API. This SDK allows developers to easily integrate fraud risk analysis into their applications.

🚀 Overview

The Fraud Detection SDK connects to a backend API that uses trained ML models to analyze transaction data and return:

• Risk Score (0–100)
• Risk Level (LOW / MEDIUM / HIGH)
• Fraud Probability

This system is designed for financial systems, blockchain analytics, and transaction monitoring applications.

✨ Features

• 🔍 Real-time fraud detection
• 📊 Risk scoring (0–100)
• ⚠ Risk level classification
• 🌐 Easy API integration
• 🧩 Lightweight and simple to use

📥 Installation

Option 1: Local Usage

git clone https://github.com/your-username/fraud-detection-sdk.git

Option 2: NPM (if published)

npm install fraud-detection-sdk

🛠 Usage

Import the SDK

import FraudDetectionSDK from "fraud-detection-sdk";

Initialize Client

const client = new FraudDetectionSDK("https://your-api-url.com");

Check Transaction

const result = await client.checkTransaction({
    blockNumber: 123456,
    confirmations: 10,
    Month: 5,
    Day: 12,
    Hour: 14,
    mean_value_received: 1.2,
    variance_value_received: 0.5,
    total_received: 10.5,
    time_diff_first_last_received: 300,
    total_tx_sent: 20,
    total_tx_sent_malicious: 2,
    total_tx_sent_unique: 15,
    total_tx_sent_malicious_unique: 1,
    total_tx_received_malicious_unique: 0
});

console.log(result);

📊 Example Response

{
  "riskScore": 87,
  "riskLevel": "HIGH",
  "fraudProbability": 0.87,
  "prediction": 1
}

📐 API Requirements

Your backend API should expose:

Endpoint:

POST /fraud-check

Request Body:

{
  "blockNumber": number,
  "confirmations": number,
  "Month": number,
  "Day": number,
  "Hour": number,
  "mean_value_received": number,
  "variance_value_received": number,
  "total_received": number,
  "time_diff_first_last_received": number,
  "total_tx_sent": number,
  "total_tx_sent_malicious": number,
  "total_tx_sent_unique": number,
  "total_tx_sent_malicious_unique": number,
  "total_tx_received_malicious_unique": number
}

⚙️ Methods

checkTransaction(features)

Sends transaction data to the API and returns fraud analysis.

Parameters:

• features (Object) → Transaction features

Returns:

• riskScore (Number)
• riskLevel (String)
• fraudProbability (Number)
• prediction (0 or 1)

getRiskLevel(score)

Returns risk category based on score.

| Score Range | Risk Level | | ----------- | ---------- | | 0–50 | LOW | | 51–80 | MEDIUM | | 81–100 | HIGH |

🧠 How It Works

User App → SDK → API → ML Model → Risk Score → Response

⚠️ Limitations

• Depends on backend API availability
• Accuracy depends on trained model quality
• Uses structured feature input (not raw wallet address yet)

🔮 Future Improvements

• Wallet address-based input (auto feature extraction)
• TypeScript support
• Browser extension integration
• Real-time blockchain data support

🧪 Development

npm install
node test.js

📄 License

MIT License

👨‍💻 Author

DEBASHISH SAHU

⭐ Final Note

This SDK demonstrates how machine learning models can be integrated into real-world applications through simple and scalable APIs.