LITESHIELD: Secure Containers via Lightweight, Composable Userspace μKernel...
summary
LITESHIELD claims VM-like isolation while avoiding VMs by reducing the user‑to‑host interface to 22 syscalls. It implements fast shared‑memory IPC and syscall redirection via LD_PRELOAD interception, runs a user‑space network stack (FSAC on DPDK bound to tap devices) and an Xv6‑like file service backed by mmap, totaling ~7,000 LOC for core services. Evaluation shows much lower getpid and read latency, better UDP throughput, weaker TCP, FIO and YCSB on Reddus that match or beat VM baselines while preserving compatibility with unmodified Linux binaries.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 0:41Monolithic vs micro kernels
- 2:30Microservices container challenges
- 5:18VM isolation vs performance
- 7:03Light Shield secure containers
- 9:13Non-delegable syscalls in userspace
- 11:19Kernel bypass for high-performance IO
- 12:23Legacy apps with LD preload
- 13:27Implementation and syscall interception
- 14:51Performance benchmarks
- 16:58Summary
related talks
26mEnabling Hardware-Assisted Shields for Linux Security Subsystems
38mSecure Containers in Embedded Deployments
22mLXDs: Towards Isolation of Kernel Subsystems
22mSandboxes all the way down - A hitchhiker's guide to platform...
38mOn the Way to Safe Containers
45mMaking Containers Safer
24mShielding Software From Privileged Side-Channel Attacks
34mChoose Your Shield: Evaluating Linux Security Modules for Cloud Na...
