LinuxKit is architected to be secure by default. This document intends to detail the design decisions behind LinuxKit that pertain to security, as well as provide context for future project direction.
LinuxKit uses modern kernels, and updates frequently following new releases. It is well understood that many kernel bugs may lurk in the codebase for years. Therefore, it is imperative to not only patch the kernel to fix individual vulnerabilities but also benefit from the upstream security measures designed to prevent classes of kernel bugs.
In practice this means LinuxKit tracks new kernel releases very closely, and also follows best practice settings for the kernel configuration from the Kernel Self Protection Project and elsewhere.
The LinuxKit project maintainers are actively collaborating with KSPP and it is an established priority for the project.
The LinuxKit kernel is intended to be identical to the upstream kernel - We only intend to carry patches that are on track to be upstreamed, or fix regressions or bugs and that we will upstream.
LinuxKit is not a full host operating system, as it primarily has two jobs: run containerd containers, and be secure.
As such, the system does not contain extraneous packages or drivers by default. Because LinuxKit is customizable, it is up to individual operators to include any additional bits they may require.
The core system components that we must include in LinuxKit userspace are key to security, and we believe they should be written in type safe languages, such as Rust, Go and OCaml, and run with maximum privilege separation and isolation.
The project is currently leveraging MirageOS to construct unikernels to achieve this, and that progress can be
tracked here: as of this writing, dhcp is the first such type safe program.
There is ongoing work to remove more C components, and to improve, fuzz test and isolate the base daemons.
Further rationale about the decision to rewrite system daemons in MirageOS is explained at length in this document.
For the daemons in which this is not complete, as an intermediate step they are running as containerd containers,
and namespaced separately from the host as appropriate.
LinuxKit's build process heavily leverages Docker images for packaging. Of note, all intermediate build images are referenced by digest to ensures reproducibility across LinuxKit builds. Tags are mutable, and thus subject to override (intentionally or maliciously) - referencing by digest mitigates classes of registry poisoning attacks in LinuxKit's buildchain. Certain images, such as the kernel image, will be signed by LinuxKit maintainers using Docker Content Trust, which guarantees authenticity, integrity, and freshness of the image.
Moreover, LinuxKit's build process leverages Alpine Linux's hardened userspace tools such as
Musl libc, and compiler options that include -fstack-protector and position-independent executable output. Go binaries
are also PIE.
LinuxKit runs as an initramfs and its system containers are baked in at build-time, essentially making LinuxKit immutable.
Moreover, LinuxKit has a read-only root filesystem: system configuration and sensitive files cannot be modified after boot. The only files on LinuxKit that are allowed to be modified pertain to namespaced container data and stateful partitions.
As such, access to the LinuxKit base system is limited in scope: in the event of any container escape, the attack surface is also limited because the system binaries and configuration is unmodifiable. To that end, the LinuxKit base system does not supply a package manger: containers must be built beforehand with the dependencies they require.
Once a secure LinuxKit base system has been built, it cannot be tampered with, even by malicious user containers. Even if user containers unintentionally expose themselves to attack vectors, immutability of the LinuxKit base system limits the scope of host attack.
By default, linuxkit has no login available: not on console, not via ssh, nowhere. You have the option of enabling login on console using a linuxkit/getty service container, but it is not created by default. Similarly, a linuxkit/sshd service container will start a sshd for you. See the getty and sshd examples.
Following the principle of least privilege for immutable infrastructure, LinuxKit cannot have the ability or attack surface to update itself. It is the responsibility of an external system, most commonly infrakit, to provision and update LinuxKit nodes.
It is encouraged to consider the notion of "reverse uptime" when deploying LinuxKit - because LinuxKit is immutable, it should be acceptable and encouraged to frequently redeploy LinuxKit nodes.
LinuxKit cannot make any trusted hardware assumptions because of the vast variety of platforms it boots on, but Infrakit can be used to provide trusted boot information and integrate with existing trusted boot hardware. In this sense, LinuxKit is "trusted boot-ready" and the team is already collaborating with cloud and hardware providers to make this a reality.
Since LinuxKit is meant to only run containers and be secure, it is the perfect platform to incubate new (and potentially radical!) paradigms and strategies for securing the Linux kernel - allowing them to be used in production environments and attract critical mass before eventually being upstreamed.
In this spirit, the /projects subdirectory houses a number of such projects. At this time, these include:
- WireGuard: a modern and minimal VPN implemented with the state-of-the-art cryptography like the Noise protocol framework
- okernel: a mechanism to split the kernel into inner and outer subkernels with different trust properties
The LinuxKit community welcomes new security projects - please propose a new project if you have one you'd like to include!