Skip to content

This solution provides an automated, serverless way to redact sensitive data from PDF files using Google Cloud Services like Data Loss Prevention (DLP), Cloud Workflows, and Cloud Run.

License

Notifications You must be signed in to change notification settings

GoogleCloudPlatform/dlp-pdf-redaction

Repository files navigation

Solution Guide

This solution provides an automated, serverless way to redact sensitive data from PDF files using Google Cloud Services like Data Loss Prevention (DLP), Cloud Workflows, and Cloud Run.

Solution Architecture Diagram

The image below describes the solution architecture of the pdf redaction process.

Architecture Diagram

Workflow Steps

The workflow consists of the following steps:

  1. The user uploads a PDF file to a GCS bucket
  2. A Workflow is triggered by EventArc. This workflow orchestrates the PDF file redaction consisting of the following steps:
    • Split the PDF into single pages, convert pages into images, and store them in a working bucket
    • Redact each image using DLP Image Redact API
    • Assemble back the PDF file from the list of redacted images and store it on GCS (output bucket)
    • Write redacted quotes (findings) to BigQuery

Deploy PDF Redaction app

The terraform folder contains the code needed to deploy the PDF Redaction application.

What resources are created?

Main resources:

  • Workflow
  • CloudRun services for each component with its service accounts and permissions
    1. pdf-spliter - Split PDF into single-page image files
    2. dlp-runner - Runs each page file through DLP to redact sensitive information
    3. pdf-merger - Assembles back the pages into a single PDF
    4. findings-writer - Writes findings into BigQuery
  • Cloud Storage buckets
    • Input Bucket - bucket where the original file is stored
    • Working Bucket - a working bucket in which all temp files will be stored as throughout the different workflow stages
    • Output Bucket - bucket where the redacted file is stored
  • DLP template where InfoTypes and rules are specified. You can modify the dlp.tf file to specify your own INFO_TYPES and Rule Sets (refer to terraform documentation for dlp templates)
  • BigQuery dataset and table where findings will be written

How to deploy?

The following steps should be executed in Cloud Shell in the Google Cloud Console.

1. Create a project and enable billing

Follow the steps in this guide.

2. Get the code

Clone this github repository go to the root of the repository.

git clone https://github.com/GoogleCloudPlatform/dlp-pdf-redaction
cd dlp-pdf-redaction

3. Build images for Cloud Run

You will first need to build the docker images for each microservice.

PROJECT_ID=[YOUR_PROJECT_ID]
PROJECT_NUMBER=$(gcloud projects list --filter="PROJECT_ID=$PROJECT_ID" --format="value(PROJECT_NUMBER)")
REGION=us-central1
DOCKER_REPO_NAME=pdf-redaction-docker-repo

# Enable required APIs
gcloud services enable cloudbuild.googleapis.com artifactregistry.googleapis.com --project $PROJECT_ID
# Create a Docker image repo to store apps docker images
gcloud artifacts repositories create $DOCKER_REPO_NAME --repository-format=docker --description="PDF Redaction Docker Image repository" --project $PROJECT_ID --location=$REGION
# Grant Required roles to the CloudBuild service account
gcloud projects add-iam-policy-binding $PROJECT_ID \
  --member="serviceAccount:[email protected]" \
  --role="roles/cloudbuild.serviceAgent"
# Build app docker images
gcloud builds submit --config ./build-app-images.yaml --substitutions _REGION=$REGION,_DOCKER_REPO_NAME=$DOCKER_REPO_NAME --project $PROJECT_ID

Note: If you receive a pop-up for permissions, you can authorize gcloud to request your credentials an make a GCP API call.

The above command will build 4 docker images and push them into Google Container Registry (GCR). Run the following command and confirm that the images are present in GCR.

gcloud artifacts docker images list $REGION-docker.pkg.dev/$PROJECT_ID/$DOCKER_REPO_NAME

4. Deploy the infrastructure using Terraform

This terraform deployment requires the following variables.

  • project_id = "YOUR_PROJECT_ID"
  • region = "YOUR_REGION_REGION"
  • docker_repo_name = "DOCKER_REPO_NAME"
  • wf_region = "YOUR_WORKFLOW_REGION"

From the root folder of this repo, run the following commands:

export TF_VAR_project_id=$PROJECT_ID
export TF_VAR_region=$REGION
export TF_VAR_wf_region=$REGION
export TF_VAR_docker_repo_name=$DOCKER_REPO_NAME

terraform -chdir=terraform init
terraform -chdir=terraform apply -auto-approve

Note: Region and Workflow region both default to us-central1. If you wish to deploy the resources in a different region, specify the region and the wf_region variables (ie. using TF_VAR_region and TF_VAR_wf_region). Cloud Workflows is only available in specific regions, for more information check the documentation.

5. Take note of Terraform Outputs

Once terraform finishes provisioning all resources, you will see its outputs. Please take note of input_bucket and output_bucket buckets. Files uploaded to the input_bucket bucket will be automatically processed and the redacted files will be written to the output_bucket bucket. If you missed the outputs from the firs run, you can list the outputs by running

terraform -chdir=terraform output

6. Test

Use the command below to upload the test file into the input_bucket. After a few seconds, you should see a redacted PDF file in the output_bucket.

gsutil cp ./test_file.pdf [INPUT_BUCKET_FROM_OUTPUT e.g. gs://pdf-input-bucket-xxxx]

If you are curious about the behind the scenes, try:

  • Checkout the Redacted file in the output_bucket.

    gsutil ls [OUTPUT_BUCKET_FROM_OUTPUT e.g. gs://pdf-output-bucket-xxxx]
    
  • Download the redacted pdf file, open it with your preferred pdf reader, and search for text in the PDF file.

  • Looking into Cloud Workflows in the GCP web console. You will see that a workflow execution was triggered when you uploaded the file to GCS.

  • Explore the pdf_redaction_xxxx dataset in BigQuery and check out the metadata that was inserted into the findings table.