Nicolai Ort
46b06c66fd
All checks were successful
Build latest image / build-container (push) Successful in 49s
58 lines
2.0 KiB
Markdown
58 lines
2.0 KiB
Markdown
---
|
|
title: What I wish i knew about container security
|
|
weight: 1
|
|
tags:
|
|
- rejekts
|
|
- security
|
|
---
|
|
|
|
<!-- {{% button href="https://youtu.be/rkteV6Mzjfs" style="warning" icon="video" %}}Watch talk on YouTube{{% /button %}} -->
|
|
<!-- {{% button href="https://docs.google.com/presentation/d/1nEK0CVC_yQgIDqwsdh-PRihB6dc9RyT-" style="tip" icon="person-chalkboard" %}}Slides{{% /button %}} -->
|
|
|
|
## BAseline
|
|
|
|
- Linux is like a hammer and containers look a lot like nails
|
|
- Containers aren't real: They are just processes with besser isolation
|
|
- IPTables is complicates
|
|
|
|
### Hard parts
|
|
|
|
- The kernel is shared we only predent to seperate processes through namespaces
|
|
- Filesystems: Containers bring a bunch of filesystems and sharing filesystems between multiple containers
|
|
- Softlinks are hard to do right because they point to a path and not the data itself
|
|
|
|
### How did we get here?
|
|
|
|
1. Unix with a buch of tools we still use
|
|
2. Linux (originally designed to for the desktop)
|
|
3. Kernel gets iptables
|
|
4. The rist concept of namespaces
|
|
5. More hypervisor stuff and official user namespaces
|
|
6. Containers (first lxc then docker)
|
|
|
|
## Sandboxing
|
|
|
|
- In browsers: They must protect the user from malicious content
|
|
- In containers: PRetty much the same - both run untrusted code that has to be isolated
|
|
|
|
## Namespaces
|
|
|
|
- Better isolation from other processes including resource constraints
|
|
- But: The shared kernel interacts with all processes (so kernel bugs can affect all namespaces)
|
|
|
|
|
|
|
|
|
|
## Improvements
|
|
|
|
- Secure Computing: Implement a secure state that we transition into before the process actually does stuff
|
|
- Paravirtualization: Instead of system calls to a shared kernel we make hyper-calls to the hypervisor
|
|
- Virtualization: The classic virtualization where everyone hosts their own kernel
|
|
|
|
## Stuff to look out for
|
|
|
|
> More or less a bit of advertisement
|
|
|
|
- Edera: Container native hypervisor without a shared kernel
|
|
- Styrolite: Rust-based container runtime sandbox
|
|
- eBPF and Tetragon for prevention and monitoring |