FraudGuard — AI Fraud Detection (Microservices Architecture)

. A high-performance fraud detection system built with a microservices architecture, AWS integration, and real-time AI-powered scoring.

Live Demo: [https://fraud-detection-theta-two.vercel.app]

🏗 Architecture

graph TD
    Client[React Dashboard / Vercel] -->|HTTP/REST| APIGateway(API Gateway)
    
    APIGateway -->|Auth Routes| AuthService(Auth Service)
    APIGateway -->|Txn Routes| TxnService(Transaction Service)
    APIGateway -->|WebSocket Auth| NotifService(Notification Service)
    
    AuthService --> DB[(MongoDB Atlas)]
    TxnService --> DB
    
    TxnService -->|Internal REST| AIService(AI Scoring Service)
    
    AIService -->|Fraud Analysis| Gemini[Google Gemini API]
    Gemini -->|Score & Reasons| AIService
    
    AIService --> DB
    AIService -->|Internal Audit| NotifService
    
    NotifService -->|Socket.io Real-time| Client
    
    Webhook[Razorpay Webhook] -->|Payment Data| APIGateway

    classDef aws fill:#FF9900,stroke:#232F3E,stroke-width:2px,color:black;
    classDef frontend fill:#61DAFB,stroke:#20232A,stroke-width:2px,color:black;
    classDef db fill:#47A248,stroke:#232F3E,stroke-width:2px,color:white;
    classDef external fill:#EA4335,stroke:#232F3E,stroke-width:2px,color:white;
    
    class APIGateway,AuthService,TxnService,AIService,NotifService aws;
    class Client frontend;
    class DB db;
    class Gemini,Webhook external;

Loading

📸 Screenshots

🧩 Services Overview

The backend is composed of modular microservices communicating via HTTP and Amazon SQS queues.

Service Name	Tech Stack	Port	Purpose
API Gateway	Node, Express, Proxy	8080	Single entry point, proxies requests to microservices, handles CORS
Auth Service	Node, Express, MongoDB	3001	User registration, login, JWT auth
Transaction Service	Node, Express, MongoDB	3002	CRUD for transactions, webhook handling (Razorpay), pushes to SQS
AI Scoring Service	Node, Gemini API	3003	Async worker, scores transactions for fraud via Google Gemini AI
Notification Service	Node, Express, Socket.io	3004	Audit logs and real-time dashboard notifications (WebSockets)

🛠 Tech Stack

Category	Technologies
Frontend	React.js (Vite), Tailwind CSS, Redux Toolkit, Recharts
Backend	Node.js, Express.js
Database	MongoDB Atlas
Cloud & DevOps	AWS (EC2, SQS, S3, CloudFront), Docker, NGINX
AI	Google Gemini API
CI/CD	GitHub Actions
Payments Integration	Razorpay Webhooks

🚀 Local Setup

1. Clone the repository

   git clone https://github.com/yourusername/FraudGuard.git
   cd FraudGuard

2. Set up Environment Variables Create a .env file in each service directory with the required keys:

   • API Gateway (services/api-gateway/.env): PORT=8080, AUTH_SERVICE_URL, TRANSACTION_SERVICE_URL, NOTIFICATION_SERVICE_URL
   • Auth Service (services/auth-service/.env): PORT=3001, MONGO_URI, JWT_SECRET
   • Transaction Service (services/transaction-service/.env): PORT=3002, MONGO_URI, SQS_QUEUE_URL, RAZORPAY_KEY_ID, RAZORPAY_KEY_SECRET
   • AI Scoring Service (services/ai-scoring-service/.env): PORT=3003, MONGO_URI, SQS_QUEUE_URL, GEMINI_API_KEY
   • Notification Service (services/notification-service/.env): PORT=3004, MONGO_URI, SQS_QUEUE_URL
   • Frontend (client/.env): VITE_API_URL=http://localhost:8080

3. Install Dependencies and Run You can run the microservices using Docker Compose (if configured) or individually by running npm install and npm run dev in each service's directory.

   For Frontend:

   cd client
   npm install
   npm run dev

4. Access the application

   • Frontend: http://localhost:5173
   • API Gateway: http://localhost:8080

🔄 CI/CD Pipeline

The project uses GitHub Actions for automated deployments:

• Backend Services: Pushes to the prod branch trigger building Docker images, pushing to a container registry, and triggering a rolling update on the AWS EC2 instance.
• Frontend: Handled automatically via Vercel or S3/CloudFront deployments on commit.

🔮 Future Improvements

• Implement a caching layer with Redis to optimize frequent queries.
• Add multi-factor authentication (MFA) for analyst accounts.
• Introduce an automated retry mechanism for failed AI scoring attempts.
• Scale services individually based on load using Kubernetes.

👨‍💻 Author

Keerthi Kumar V LinkedIn | GitHub