SourceAWS Machine Learning Blog•Stalecollected in 6m
Huntington Bank scales PII redaction with AWS
Learn how to automate sensitive data redaction at a 400M+ document scale using AWS ML pipelines.
30-Second TL;DR
What Changed
Processed 400 million+ documents for sensitive data redaction
Why It Matters
Demonstrates the viability of large-scale automated compliance and data governance using cloud-native machine learning pipelines.
What To Do Next
Review your data compliance workflows and evaluate AWS ML services for automating PII redaction in high-volume document processing.
Who should care:Enterprise & Security Teams
Key Points
- •Processed 400 million+ documents for sensitive data redaction
- •Achieved 95%+ redaction accuracy using AWS ML services
- •Reduced document processing lifecycle from years to months
Deep Insight
AI-generated analysis for this event — not the original article.
Enhanced Key Takeaways
- •The solution leverages Amazon Comprehend to identify and redact Personally Identifiable Information (PII) and Payment Card Industry (PCI) data across unstructured document repositories.
- •Huntington Bank utilized a serverless architecture, specifically AWS Lambda and Amazon S3, to handle the massive scale of document ingestion and processing without managing underlying infrastructure.
- •The implementation was part of a broader digital transformation strategy aimed at enhancing data governance and regulatory compliance while migrating legacy document archives to the cloud.
- •The project incorporated a human-in-the-loop (HITL) review process for low-confidence redactions to ensure the 95% accuracy threshold was maintained and improved over time.
- •By automating the redaction pipeline, Huntington Bank significantly lowered the operational cost per document compared to manual redaction or traditional on-premises OCR solutions.
Competitor Analysis
PII Redaction
- AWS (Huntington Case)
- Native Amazon Comprehend
- Google Cloud (Document AI)
- Specialized PII Redaction API
- Microsoft Azure (Form Recognizer)
- Azure AI Language PII detection
Scalability
- AWS (Huntington Case)
- High (Serverless/Lambda)
- Google Cloud (Document AI)
- High (Managed/Auto-scaling)
- Microsoft Azure (Form Recognizer)
- High (Container/Managed)
Accuracy
- AWS (Huntington Case)
- 95%+ (Reported)
- Google Cloud (Document AI)
- Varies by model/tuning
- Microsoft Azure (Form Recognizer)
- Varies by model/tuning
Integration
- AWS (Huntington Case)
- Deep AWS Ecosystem
- Google Cloud (Document AI)
- Deep GCP/Workspace
- Microsoft Azure (Form Recognizer)
- Deep M365/Power Platform
| Feature | AWS (Huntington Case) | Google Cloud (Document AI) | Microsoft Azure (Form Recognizer) |
|---|---|---|---|
| PII Redaction | Native Amazon Comprehend | Specialized PII Redaction API | Azure AI Language PII detection |
| Scalability | High (Serverless/Lambda) | High (Managed/Auto-scaling) | High (Container/Managed) |
| Accuracy | 95%+ (Reported) | Varies by model/tuning | Varies by model/tuning |
| Integration | Deep AWS Ecosystem | Deep GCP/Workspace | Deep M365/Power Platform |
Technical Deep Dive
- Architecture utilizes Amazon S3 as the primary data lake for storing raw and redacted document versions.
- Employs Amazon Comprehend PII entities detection API to scan text extracted from documents.
- Uses AWS Step Functions to orchestrate the workflow, including document splitting, text extraction, redaction, and final validation.
- Implements Amazon Textract for high-fidelity OCR to convert scanned PDFs and images into machine-readable text before redaction.
- Employs AWS Identity and Access Management (IAM) and AWS Key Management Service (KMS) to ensure data encryption and strict access control during the processing lifecycle.
Future ImplicationsAI analysis grounded in cited sources
Financial institutions will shift toward real-time PII redaction for all incoming digital documents.
The success of batch processing at this scale proves that automated redaction is reliable enough to be integrated into live customer-facing workflows.
Generative AI will replace traditional entity-based redaction models within 24 months.
LLMs offer superior context awareness for identifying sensitive data in complex, non-standardized legal and financial documents compared to traditional NER models.
Timeline
2021-05
Huntington Bank completes acquisition of TCF Financial, necessitating massive data integration.
2022-11
Huntington Bank accelerates cloud migration strategy to modernize legacy document storage.
2024-03
AWS highlights Huntington Bank's scalable PII redaction architecture in technical case study.
- 2021-05Huntington Bank completes acquisition of TCF Financial, necessitating massive data integration.
- 2022-11Huntington Bank accelerates cloud migration strategy to modernize legacy document storage.
- 2024-03AWS highlights Huntington Bank's scalable PII redaction architecture in technical case study.
Weekly AI Recap
Read this week's curated digest of top AI events →
Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: AWS Machine Learning Blog ↗
This is a summary, not the original. Read the source, or get the weekly briefing.
The weekly digest
One email a week. Unsubscribe anytime.