Security for AI and Multi-Party Collaboration with Confidential Computing (Oh, and...
summary
Describes how hardware Trusted Execution Environments produce a cryptographic attestation measurement (CPU/GPU signs a hash with an unextractable key) and how that enables five concrete AI guarantees including provenance of inference back to training data and model, a three‑bank reconciliation example that preserves each bank's confidentiality, a pharma workflow where hospitals keep raw patient records private while audited code runs on the data, and a Web3 pattern combining smart contracts, DApp hashes and attested resource offers with an implementation by Super Protocol and CCC work on verification and n100 GPU TEEs.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:04Problem with abstraction
- 2:47What is confidential computing
- 3:31Attestation in security
- 6:04AI use case
- 9:54Multi-party finance
- 13:11Pharmaceutical research
- 18:36Web 3 smart contracts
- 25:59Q&A on use cases
- 29:11Confidential Computing Consortium
- 35:41Q&A and closing
related talks
42mSponsored Session: Confidential Computing - New Capabilities for New Workloads- 42mSponsored Session: Confidential Computing - New Capabilities for New Workloads
37mIt's All About Trust: Why Confidential Computing Matters
44mGPU-less, Trust-less, Limit-less: Reimagining the Confidential AI Cloud- 37mIt's All About Trust: Why Confidential Computing Matters
20mTrusted Hardware for AI Workloads: Extending Confidential...
19mHow Can We Have Private Access to Intelligence?
39m"I Didn't Peek: and I Can Prove It": Confidential Computing for Audits and Regulators