Beyond Labels: Evolving Data Classification
summary
Roblox runs a risk‑first data classification program for ~72 million daily users and 2.5M developers, aiming to scale to one billion; it prioritizes scope/stratify/secure, uses decision trees, DLP discovery, metadata tagging, access controls, encryption and phased automation, and ties ownership to VP levels. Examples: AI‑generated slide art and a past program that reduced categories from 48 to 6 to improve adoption. Claims: right‑sized controls beat perfect design, classify by data type then size, and embed taxonomy into workflows so labels enable enforcement rather than just awareness.
AI-written summary · about
chapters (6)
- 0:00Introduction and context
- 1:25Picking the right framework
- 8:03Scope stratify and secure
- 18:22Securing classified data
- 23:53Ongoing management and alignment
- 25:35Q&A session
related talks
30mData Driven Security
50mNotes from the field Microsoft Information Protection
05mDesigning Toxic Content Classification for a Diversity of Perspectives
12mHow Predatory Monetization Designs Manifest in Child-Friendly Video Games
01h 01mD1T2 - Applying Machine Learning to User Behavior Anomaly Analysis
31mStandard Classification: Live Q&A Session
24mData’s Rotten Scoundrels – How To Avoid Becoming One
11mCreating Effective Labels