Demystifying Privacy Preserving Computing
summary
Survey of concrete PPC techniques and tradeoffs with deployment notes and numbers so you can pick by what matters: reviews Federated learning for edge training and its parameter leakage, explains differential privacy with epsilon and Apple’s unreported epsilon leaking privacy, walks through secure multi‑party computation and Shamir secret sharing for threshold safety, and covers homomorphic encryption plus performance reality (IBM HElib 2016 called “trillion×” slower and now ~75×). Mentions real uses (US Census, CryptoNets, Microsoft libraries) and practical bottlenecks.
AI-written summary · about
chapters (10)
- 0:00Start
- 1:11Introduction
- 5:48Impact of Covid19 on lungs?
- 9:40Federated Learning
- 14:59Uses of Differential privacy
- 16:26Secure multi-party computation
- 18:30Blueprint for MPC
- 18:43Splitting into secrets
- 22:34Homomorphic Encryption
- 27:13What does the future hold?
related talks
38mDemystifying Privacy Preserving Computing
29mPrivacy-preserving methods: Building secure projects
38mApplying Privacy-Enhancing Technologies
37mBoF: PETs - Privacy-Enhancing Technologies Compared, Discussed &...
12mLessons from years of building privacy-preserving technologies
39mData Privacy Engineering to power Responsible Enterprise AI
40mConfidential Computing and Privacy-Enhancing Technologies - The Landscape
01h 01mToward Provably Private Federated Learning