FraudBuster - Fraud Detection & Review Analysis System

A cloud-distributed microservices system for detecting fraudulent and bot-generated product reviews using machine learning models, gRPC communication, and distributed caching.

📋 Overview

FraudBuster analyzes Amazon product reviews to identify generic, bot-generated, and fraudulent reviews using pretrained BERT models. The system features a microservices architecture with 6 custom services communicating via gRPC, backed by Redis caching and MongoDB persistence.

🏗️ Architecture

Microservices:

• flask-app - HTTP API Gateway & Web Frontend (Port 5000)
• review-classifier - ML-powered review classification using BERT (Port 50051)
• relevance-scorer - Review relevance scoring service (Port 50052)
• scraper - Amazon product review scraper (Port 50053)
• mongo-store - gRPC database writer service (Port 50054)
• scheduler - Background task orchestrator

Infrastructure:

• Redis - Distributed cache layer (Port 6379)
• MongoDB 7 - Persistent data storage (Port 27017)

Communication: All microservices communicate via gRPC with Protocol Buffers

📁 Project Structure

.
├── dataset/                    # Training dataset for review-classifier model
├── flask-app/                  # API Gateway microservice
├── k8s/                        # Kubernetes deployment files
│   ├── run_fraudbuster(win).bat    # Windows deployment script
│   └── run_fraudbuster(mac).sh     # Mac/Linux deployment script
├── mongo-store/                # Database writer microservice
├── protos/                     # gRPC Protocol Buffer definitions
├── relevance-scorer/           # Review scoring microservice
├── review-classifier/          # ML classification microservice
├── scheduler/                  # Task orchestration microservice
├── scraper/                    # Web scraping microservice
├── docker-compose.yml          # Docker Compose configuration
└── README.md                   # This file

🚀 Quick Start

Prerequisites

• Docker Desktop (version 20.10+)
• Docker Compose (version 2.0+)
• (Optional) Minikube + kubectl for Kubernetes deployment

Installation

1. Clone the repository:

   git clone <repository-url>
   cd <repository-folder>

2. Pull all service images from Docker Hub:

   docker-compose pull

   Wait time: ~2-5 minutes depending on internet speed

3. Start all services:

   docker-compose up --build

   Services will initialize in ~30 seconds

4. Verify all containers are running:

   docker ps

   Expected: 8 running containers

5. Access the application:

   http://localhost:5000
   

Basic Commands

View logs:

docker-compose logs -f <service-name>
# Example: docker-compose logs -f flask-app

Restart a specific service:

docker-compose restart <service-name>

Stop all services:

docker-compose down

Update to latest images:

docker-compose pull
docker-compose up -d --force-recreate

☸️ Kubernetes Deployment (Alternative)

For Kubernetes/Minikube deployment:

1. Ensure Minikube is running:

   minikube start

2. Navigate to k8s folder:

   cd k8s

3. Run the deployment script:

   Windows:

   run_fraudbuster(win).bat

   Mac/Linux:

   chmod +x run_fraudbuster(mac).sh
   ./run_fraudbuster(mac).sh

4. Check pod status:

   kubectl get pods -w

   Wait until all pods show STATUS: Running

5. Access the service:

   minikube service flask-app-service

🧪 Testing

For comprehensive testing instructions, refer to the GROUPNN_INSTRUCTIONS.pdf file included in the submission.

The PDF contains:

• Detailed testing scenarios
• Expected outputs for each test case
• Troubleshooting guide
• Performance benchmarks

🛠️ Built With

• Backend: Flask, Python 3.10+
• ML Models: BERT (Pretrained Transformers)
• Communication: gRPC, Protocol Buffers
• Databases: MongoDB 7, Redis
• Orchestration: Docker Compose, Kubernetes
• Container Registry: Docker Hub

📦 Docker Images

All microservice images are hosted on Docker Hub:

• caffeinatedkong/flask-app:latest
• caffeinatedkong/review-classifier:latest
• caffeinatedkong/relevance-scorer:latest
• caffeinatedkong/scraper:latest
• caffeinatedkong/mongo-store:latest
• caffeinatedkong/scheduler:latest

Public repositories available at: https://hub.docker.com/u/caffeinatedkong

🔧 Configuration

All services are pre-configured via docker-compose.yml. No manual configuration required for basic deployment.

Environment Variables:

• PYTHONUNBUFFERED=1 - Real-time Python logging
• Service hostnames auto-configured via Docker network

📊 Service Ports

Service	Port	Protocol
flask-app	5000	HTTP
review-classifier	50051	gRPC
relevance-scorer	50052	gRPC
scraper	50053	gRPC
mongo-store	50054	gRPC
redis	6379	TCP
mongo	27017	TCP

🐛 Troubleshooting

Port conflicts:

# Check if ports are in use
netstat -an | findstr "5000"

Services not starting:

# Check service logs
docker-compose logs <service-name>

Old cached images:

docker-compose down
docker rmi -f caffeinatedkong/<service-name>:latest
docker-compose pull
docker-compose up -d

📝 Project Information

Course: CSC3104 - Cloud and Distributed Computing
Institution: Singapore Institute of Technology
Submission Date: November 16, 2025

👥 Contributors

Group 09 - FraudBuster Team

📄 License

This project is submitted as part of academic coursework.

---

For detailed testing instructions and submission requirements, please refer to GROUP09_INSTRUCTIONS.pdf.