redaction

**Data redaction** is the **process of automatically detecting and removing sensitive information from text, documents, and datasets** — using NLP and pattern matching to identify PII (personally identifiable information), credentials, financial data, and protected health information before sharing, storing, or processing data. **What Gets Redacted** - **PII**: Names, addresses, phone numbers, email addresses, SSNs. - **Financial**: Credit card numbers, bank accounts, salary data. - **Health (PHI)**: Medical records, diagnoses, treatment information. - **Credentials**: API keys, passwords, tokens, connection strings. - **Legal**: Attorney-client privileged communications. **Redaction Methods** - **Pattern Matching**: Regex for structured data (SSN: XXX-XX-XXXX). - **NER (Named Entity Recognition)**: ML models detect names, locations, organizations. - **LLM-Based**: Use language models to identify contextual sensitive information. - **Tokenization**: Replace sensitive values with non-reversible tokens. **Tools**: Microsoft Presidio, AWS Comprehend PII, Google DLP API, spaCy NER. Data redaction is **essential for privacy compliance** — enabling organizations to share and process data safely while meeting GDPR, HIPAA, and CCPA requirements.