Fraud Detection in GCP

This repository is used to implement an event-based fraud detection system using Infrastructure as Code. The infrastructure is fully serverless and includes the following GCP services:

• BigQuery
• Cloud Build
• Cloud Functions
• Cloud Pub/Sub
• Cloud Run
• Cloud Storage
• Firestore

Prerequisites

To run the demo, you need to have a Google Cloud project (that has the APIs listed above enabled), and also have Terraform installed on your device. Please stick to the steps below to set up ever

Steps

• Install the gcloud CLI
• Install the Terraform CLI
• Set $GCP_CREDENTIALS and $GCP_PROJECT_ID as GitHub Actions secrets
• Authenticate with your GCP project using gcloud auth application-default login
• Push the image to your Artifact Repository

gcloud builds submit --tag <YOUR_GCP_REGION>-docker.pkg.dev/<YOUR_GCP_PROJECT_ID>/fraud-detection/event-producer`

Running the demo

1. Set environment variables

export TF_VAR_region=<YOUR_GCP_REGION>
export TF_VAR_project_id=<YOUR_GCP_PROJECT_ID>

2. Change the Terraform state file bucket

event-driven-fraud-detection/infrastructure/main.tf

Line 3 in bf10748

bucket = "tm-fraud-detection-tfstate"

3. Initialize the Terraform backend

cd infrastructure
terraform init

4. Run Terraform

terraform apply

5. Run the application

• Populate the BigQuery dataset with data (most importantly the venues and users tables)
• Go to the Cloud Run console and use the generated URL to access the producer service
• Use the Streamlit UI to produce events

6. View the results

• After a certain number of events have been generated, you can check the fraud table to see which checkins have been deemed as fraudulent and need further investigation. Furthermore, you may choose to implement an analytics tool such as Looker Studio to produce dashboards and visualize your findings.

Destroy the infrastructure

terraform destroy