Kubernetes + Encrypted Memory = Security * Privacy
summary
Hardware-backed memory encryption plus a VM runtime protects in-use secrets by running containers inside an IBM secure VM (SVM) launched via the kata container runtime. The team moves Kubernetes EmptyDir ephemeral volumes and an encrypted TensorFlow model entirely inside SVM memory so the host root sees only encrypted pages; demo inference returns correct labels (Persian cat) while the model file is invisible on the node. Remaining gaps: host-side image extraction and ongoing KEP work for encrypted container images.
AI-written summary · about
chapters (12)
- 0:00Start
- 0:23Who Are We?
- 1:21How Do We Secure Data and Code?
- 3:01What is being done?
- 4:33Explain Like I'm 5
- 5:09Secure VM and Secure Containers
- 9:08How can we use it with k8s?
- 11:08Secure Containers with Kata
- 11:41Kata Containers
- 20:42Demo
- 21:48Work in Progress
- 22:49References
related talks
29mConfidential Containers: The Next Era of Cloud Data Security
05mSponsored Keynote: Bringing Confidential Computing to Kubernetes
38mConfidential Containers Explained
46mConfidential Cloud: Safeguarding Sensitive Data in K8s
33mDon't Peek Into my Container!
38mCloud native Confidential: Isolating Kubernetes Pods with Confidential Computing
28mPreserving Data and AI/ML Model Privacy Using Confidential..
38mHardware-based KMS Plug-in to Protect Secrets in Kubernetes