Real-Time Fraud Detection Service

Overview

This project implements a real-time fraud detection backend service using Spring Boot, Kafka, and Redis. It processes user transactions asynchronously via Kafka, caches recent transactions in Redis to enable temporal validation, and provides real-time status updates to clients using Server-Sent Events (SSE).

Features

• Asynchronous transaction processing via Kafka topics.
• Redis based caching of recent user transactions with automatic expiry.
• Transaction spam prevention by allowing 1 transaction every 5 seconds.
• Rate limiting based on client IP and session.
• Real-time transaction status updates via Kafka and SSE.
• Session management and client IP tracking for enhanced fraud prevention.
• Geolocation to calculate distance between transactions using haversine formula
• Fraud prediction using binary logistic regression model using features such as amount, location, time, and transaction frequency.
• Smile binary logistic regression library to predict if a transaction

Tech stack

• Java 17+
• Spring Boot
• Apache Kafka
• Redis via Spring Data Redis
• Jackson JSON serialization
• Server side events for realtime updates
• Smile ML
• Geolocation tools

Prerequisites

• Java Development Kit (JDK) 17 or later
• Apache Kafka (https://kafka.apache.org/downloads)
• Redis Server (sudo apt install redis-server) or through App center if you’re on Ubuntu Linux
• Maven or Gradle build tool

Kafka setup

Go into your Kafka directory

cd ~/kafka/kafka_2.13-3.9.1

Generate a UUID

bin/kafka-storage.sh random-uuid

Format the storage using your UUID

bin/kafka-storage.sh format -t <YOUR_UUID_HERE> -c config/kraft/server.properties

Start the Kafka server

bin/kafka-server-start.sh config/kraft/server.properties

Redis setup

Run redis-cli in bash

Application.properties

spring.application.name=FraudDetector
spring.kafka.bootstrap-servers=localhost:9092
spring.cache.type=redis
spring.data.redis.host=localhost
spring.data.redis.port=6379
logging.level.org.springframework.cache=DEBUG
logging.level.org.springframework.data.redis.cache=DEBUG

management.endpoints.web.exposure.include=*
management.endpoint.health.show-details=always

spring.jackson.serialization.write-dates-as-timestamps=false

How It Works

• Client submits transactions via /api/tr.
• The transaction is sent to Kafka "transactions" topic asynchronously.
• TransactionService consumes transactions from Kafka, retrieves recent transactions from Redis, and validates timing to prevent rapid repeats.
• Status updates are sent back on the "out-transactions" Kafka topic keyed by session ID.
• Clients receive live updates via SSE from /api/streams/results.
• Redis caches recent transactions per user/session with a 10-minute expiry that refreshes on each new transaction.
• ContributingContributions are welcome.
• Please fork the repository and submit pull requests with clear descriptions.