Distributed Transaction & Fraud Detection System

A microservices-based digital banking platform implementing the SAGA pattern for distributed transaction consistency, with real-time fraud detection and event-driven notifications.

Services Overview

Service	Port	Responsibility
api-gateway	8080	Single entry point, rate limiting
account-service	8081	Account management, balance, deduct/credit/refund
transaction-service	8082	SAGA orchestrator — money transfers, transaction history
payment-service	8083	Razorpay integration, external payments
fraud-detection-service	8084	Real-time fraud detection via Redis
notification-service	8085	Transaction and fraud alerts (email/SMS)

Architecture Flow

User → API Gateway (rate limiting)
             ↓
      Transaction Service (SAGA orchestrator)
             ↓
        Account Service
     (validate balance, deduct)
             ↓
        Apache Kafka
             ↓
    ┌────────────────────────┐
    │                        │
Fraud Detection      Transaction Service
(Redis patterns)     (decides next step)
    │                        │
    ▼                        ▼
Account Service        Notification Service
(block if fraud /      (debit/credit/fraud alerts)
 credit receiver)

Why SAGA?

Each service owns its own database (account DB, transaction DB, payment DB). A traditional single-database system rolls back automatically on failure — in a distributed setup, no single database can undo a step that already committed in another service. The transaction-service acts as the SAGA orchestrator: it tracks each step (deduct → fraud-check → credit) and triggers explicit compensating actions (e.g. refund, unblock) if a later step fails, rather than relying on an implicit rollback.

Kafka Topics

Topic	Publisher	Consumer
transaction.initiated	Transaction Service	Fraud Detection Service
fraud.check.result	Fraud Detection Service	Transaction Service
transaction.completed	Transaction Service	Account Service, Notification Service
fraud.detected	Fraud Detection Service	Account Service, Notification Service
payment.completed	Payment Service	Notification Service

Tech Stack

• Backend: Java, Spring Boot, Spring Cloud Gateway, OpenFeign
• Messaging: Apache Kafka
• Caching: Redis (fraud pattern detection, rate limiting)
• Database: MySQL (per-service, separate schemas)
• Payments: Razorpay integration
• Infra: Docker (Kafka, Zookeeper, Redis, MySQL via docker-compose)

How To Run

Step 1: Start Infrastructure

docker-compose up -d

This brings up Kafka, Zookeeper, Redis, and MySQL (skip MySQL in docker-compose if you're running it locally already).

Step 2: Start All Services

# Terminal 1
cd account-service && mvn spring-boot:run

# Terminal 2
cd transaction-service && mvn spring-boot:run

# Terminal 3
cd payment-service && mvn spring-boot:run

# Terminal 4
cd fraud-detection-service && mvn spring-boot:run

# Terminal 5
cd notification-service && mvn spring-boot:run

# Terminal 6
cd api-gateway && mvn spring-boot:run

Step 3: Configure Credentials

Before testing, set the following in each service's application.yml:

• MySQL username/password
• Razorpay API key/secret (payment-service)
• Email/SMS credentials (notification-service)

Note: Do not commit real credentials. Use environment variables or a .gitignore'd local properties file.

Example Flow

1. User initiates a transfer → POST /api/v1/transactions/transfer
2. API Gateway rate-limits and forwards to Transaction Service
3. Transaction Service calls Account Service to validate and deduct balance
4. Transaction saved with status PROCESSING; transaction.initiated published to Kafka
5. Fraud Detection Service consumes the event, checks Redis-based patterns
6. If clean → Transaction Service credits receiver, publishes transaction.completed
7. If suspicious → account is flagged/blocked, fraud.detected published, user notified
8. Notification Service sends alerts at each key step (debit, credit, fraud, block)