Open the gate a little: strategies to protect and share data
summary
Claims that you can safely give limited production data while cutting leakage risk, backed by concrete cases: an unencrypted laptop with ~20,000 PHI records triggers a >$1M HIPAA fine and a 132 GB MongoDB leak exposes 93.4 million Mexican records. Demonstrates a possible_synonymizer Postgres extension with static masking, column shuffling, noise and dynamic masking plus masked pg_dump export, shows anonymized dumps and materialized-view generalization working on a demo dataset.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:50Data protection in production
- 4:01Understanding HIPAA and protected health information
- 7:14Global data laws overview
- 8:59Exposed data consequences
- 9:20Avoid full production copies
- 10:24Anonymize data before sharing
- 12:34Setting up pgAnonymizer
- 13:38Static masking with pgAnonymizer
- 15:01Masking rules for columns
- 17:52Static masking results
- 19:38Shuffling columns for privacy
- 20:40Adding noise to numerical data
- 21:45Dynamic masking for roles
- 22:50Export masked data with pg_dump
- 23:32Generalizing medical records
- 25:18Key takeaways and closing thoughts
related talks
32mAnonymizing Data in Postgres Without Losing Your Mind (or Your
32mAnonymization and Dynamic Masking with PostgreSQL
34mDatabase Anonymization with PostgreSQL
28mDon't leak user data to AI - Strategies for protecting PII from LLMs and MCP
46mProtecting sensitive data in huge datasets: Cloud tools you can use
24mStop Making Your Data Team the 'Data Police'
31mLeverage AI to Protect Your Data Assets... from AI
50mPrivacy in Practice with Smart Pseudonymization: Lessons from the Belgian Public Sector