Protecting sensitive data in huge datasets #bbuzz
summary
Shows concrete workflows using Google DLP API, BigQuery and Data Prep to scan and transform massive public tables, detect PII in text and images, do format‑preserving encryption, masking and date shifting, and compute k‑anonymity and l‑diversity plus the newly added delta‑presence metric. Demonstrates on 170M NYC taxi trips and ~13M Mexican birth records, quantifies tradeoffs (dropping small groups removes <0.04% nationally but >1% in some states) and explains how measures change what analysis remains possible.
AI-written summary · about
chapters (16)
- 0:00Start
- 9:23Google BigQuery
- 10:31Take charge of your data with DLP API Provides capabilities
- 10:43Examples of sensitive data
- 11:26Simple example
- 13:01Risk analysis - Some definitions
- 13:37Risk analysis - Measuring
- 16:53Real datasets: Mexican births.
- 22:48K-map: additional "attack" dataset
- 24:17k-map example
- 26:11NYC taxis iterations
- 27:31Anonymity solutions
- 30:08Mask, de-ID, and transform data
- 30:24De-ID: Masking and Tokenization
- 31:10De-ID:Date Shifting
- 35:47GCP Tools
related talks
41mAccelerating privacy-enhancing data processing
39mData Privacy Engineering to power Responsible Enterprise AI
18mData Mapping at a Billion Dollar Self-Driving Vehicle Startup
01h 59mDeep Dive Workshop – Data Security and Data Protection
33mMachine Learning in the age of increasing Data Privacy Consciousness
32mThe Dangerous Data Anonymization - Keynote
08mEvil Uses of Data
10mWhen Privacy Guarantees Meet Pre-Trained LLMs: A Case Study in Synthetic Data