Automatically Securing Linux Application Containers in Untrusted Clouds
summary
Graphene implements a library OS that runs unmodified Linux apps inside Intel SGX enclaves, mapping >300 syscalls to ~40 PAL APIs and using ~50 host syscalls; it adds remote attestation and a secret‑provisioning flow tied to Intel PCS and the quoting enclave. Graphene Shielded Containers (GSC) auto‑generates manifests from Docker images, signs enclaves, and demoed a working Python/Hello print and PyTorch object‑detection workflow using a protected file system with SHA‑256 integrity. Limitations: SGXv1 enclave size, no side‑channel mitigations, and out‑of‑tree SGX drivers and Kubernetes device plumbing.
AI-written summary · about
chapters (11)
- 0:00Start
- 0:37Data protection gap in an untrusted cloud
- 3:23Extending Intel SGX Enclaves to run applications
- 7:28Graphene-SGX Manifest • To configure the application environment and security policies, developers provide a manifest file for each application
- 9:05Graphene Remote Attestation
- 14:14Graphene Protected Files
- 16:18Containers are deployment model for cloud
- 16:59Graphene Shielded Containers (GSC) in a Nutshell
- 18:08Automatically generating Graphene manifest
- 19:56GSC Cloud Deployment Scenario III
- 21:11End-to-End Use Case using Pytorch
related talks
32mSecure Container with SGX: Protecting Secret in Cloud Environment
32mSecure Container with SGX: Protecting Secret in Cloud Environment
36mAttacking & Defending Kubernetes TEE Enclaves in Critical Infrastructure
35mConfidential Computing in the Control Plane (case study using Barbican)
05mSponsored Keynote: Bringing Confidential Computing to Kubernetes
40mKubernetes Meets Confidential Computing - The Different Ways of Scaling Sensitive...
46mState of Intel SGX in Linux
18mYou Can Sign It, But Can You Trust It? Securing the Compilation Process