Preserving Data and AI/ML Model Privacy Using Confidential..
summary
Shows a working pipeline that runs a containerized LLM inside a VM-based trusted execution environment on Kubernetes using CNCF confidential-containers and a runtimeClass, with keys exchanged by a secure connect layer. An attestation + verifier service proves the workload image and TE state, and traffic to/from the pod is end-to-end encrypted so memory dumps yield garbage. Demo deploys a chatbot image on a confidential VM and verifies that encrypted requests and attestations flow correctly.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:03AI chatbot model theft
- 7:44Cloud trust risks
- 10:36What is confidential computing
- 14:09Trusted execution environments
- 16:39Trust model comparison
- 19:29Securing AI/ML workloads
- 23:01Encrypt AI with CNCF
- 24:04Confidential computing demo
- 26:30Role of attestation
related talks
29mConfidential Containers: The Next Era of Cloud Data Security
31mConfidential Containers – Sensitive Data and Privacy in Cloud Native Environments
50mConfidential Containers: Why, How, and Where Are We?
38mConfidential Containers Explained
47mTowards Confidential AI for the Masses!
46mConfidential Cloud: Safeguarding Sensitive Data in K8s
38mCloud native Confidential: Isolating Kubernetes Pods with Confidential Computing
27mTrust No One: Bringing Confidential Computing to Containers