Skip to content

Latest commit

 

History

History
186 lines (128 loc) · 10.9 KB

SECURITY_FEATURES.md

File metadata and controls

186 lines (128 loc) · 10.9 KB

Security Features

Goals

Bottlerocket has the following high-level security goals. They are listed in priority order.

  1. Harden the OS against persistent threats.
  2. Mitigate the impact of vulnerabilities in the OS.
  3. Protect containers from other containers.

We will add and enhance security features over time based on these goals.

Overview

Feature Version
Automated security updates 0.3.0
Immutable rootfs backed by dm-verity 0.3.0
Stateless tmpfs for /etc 0.3.0
No shell or interpreters installed 0.3.0
Executables built with hardening flags 0.3.0
SELinux enabled in enforcing mode 0.3.0
Kernel lockdown in integrity mode 1.1.0
Secure Boot enabled 1.15.0

The version listed indicates the first release of Bottlerocket that included the feature. Features may evolve or improve over time.

Details

Automated security updates

Bottlerocket is designed for reliable security updates that can be applied through automation.

This is achieved through the following mechanisms:

  • Two partition sets and an active/passive flip to swap OS images
  • Declarative API with modeled settings for runtime configuration
  • Variants to silo backwards-incompatible or breaking changes

Using partition sets and modeled settings removes the dependency on correct local state for reliable updates. There is no package manager database or shared filesystem tree that can become corrupted and make the process non-deterministic.

Update Policy

Our philosophy for variants is that the right time for an unexpected major version update to the kernel or orchestrator agent is "never". New variants can introduce newer LTS kernels or GPU drivers. On release, variants peg to a kernel and GPU driver version and relevant security patches are applied. However, in a situation where security patches are no longer available for the kernel or GPU driver, an existing variant may adopt a new version to address security vulnerabilities.

Kubernetes variants

Bottlerocket provides updates for each Kubernetes variant for approximately 14 months after the first release of each variant. For aws-k8s-* variants, Bottlerocket follows the Amazon EKS support policy, including extended support beyond the typical 12 months support period.

We provide a Kubernetes operator for automated updates to Bottlerocket. We recommend deploying it on your Kubernetes clusters.

ECS variant

Bottlerocket provides updates for each ECS variant for at least one year after the first release of each variant. Because the ECS agent is backwards compatible, there is no need to create new variants on a regular cadence. ECS variants will be added as necessary to introduce newer LTS kernels or potentially breaking changes.

We provide an updater for automated updates to Bottlerocket. We recommend deploying it on your ECS clusters.

Immutable rootfs backed by dm-verity

Bottlerocket uses dm-verity for its root filesystem image. This provides transparent integrity checking of the underlying block device using a cryptographic digest.

The root filesystem is marked as read-only and cannot be directly modified by userspace processes. This protects against some container escape vulnerabilities such as CVE-2019-5736.

The kernel is configured to restart if corruption is detected. That allows the system to fail closed if the underlying block device is unexpectedly modified and the node is in an unknown state. The uncontrolled reboot will disrupt running containers, which can trigger alarms and prompt administrators to investigate.

Although this provides a powerful layer of protection, it is incomplete unless Secure Boot is enabled. Otherwise, an attacker with full access to the block device could alter both the verity metadata and the contents of the root filesystem.

Stateless tmpfs for /etc

Bottlerocket uses tmpfs, a memory-backed filesystem, for /etc.

Direct modification of system configuration files such as /etc/resolv.conf or /etc/containerd/config.toml is not supported. This makes OS updates more reliable, as it is not necessary to account for local edits that might have changed the behavior of system services. It also makes it harder for an attacker to modify these files in a way that persists across a reboot.

There are two supported ways to configure the OS in the presence of these restrictions.

The first is through the API. Settings are persisted across reboot and migrated through OS upgrades. They are used to render system configuration files from templates on every boot.

The second is by using containers. Specifications such as CNI and CSI provide ways to configure networking and storage devices. Containers written to these specifications can be deployed to nodes using orchestrator-specific mechanisms like DaemonSets.

All variants will include a secondary filesystem for local storage. It will be mounted at /local with bind mounts for /var and /opt. Modifications to this area will survive an OS update or a reboot.

No shell or interpreters installed

Bottlerocket does not have a shell installed in non-developer builds. Interpreted languages such as Python are not installed or even available as packages.

Shells and interpreters enable administrators to write code that combines other programs on the system in new ways. However, these properties can also be exploited by an attacker to pivot from a vulnerability that grants local code execution.

The lack of a shell also serves as a forcing function to ensure that new code for the OS is written in a preferred language such as Rust or Go. These languages offer built-in protection against memory safety issues such as buffer overflows.

Executables built with hardening flags

The GCC cross-compilers in the Bottlerocket SDK are built with these options:

  • --enable-default-pie for -fPIE and -pie by default
  • --enable-default-ssp for -fstack-protector-strong by default

Position-independent executables (PIE) have their address space randomized on every execution. This makes addresses harder to predict for an attacker that attempts to exploit a memory corruption vulnerability.

The stack protector feature enables stack canaries to detect stack overflow and abort the program if it occurs. The "strong" version enables it for additional functions.

All C and C++ programs are compiled with the following options:

  • -Wall to warn about questionable constructs
  • -Werror=format-security to warn about unsafe uses of format functions
  • -Wp,-D_FORTIFY_SOURCE=2 for runtime error checks in libc
  • -Wp,-D_GLIBCXX_ASSERTIONS for runtime error checks in libstdc++
  • -fstack-clash-protection for stack overflow detection

Although C and C++ lack the memory safety of Go and Rust, these options add a layer of defense during build and execution.

All binaries are linked with the following options:

  • -Wl,-z,relro to mark segments read-only after relocation
  • -Wl,-z,now to resolve all symbols at load time

Together these enable full RELRO support which makes ROP attacks more difficult to execute.

Note: Certain variants, such as the ones for NVIDIA, include precompiled binaries that may not have been built with these hardening flags.

SELinux enabled in enforcing mode

Bottlerocket enables SELinux by default, sets it to enforcing mode, and loads the policy during boot. There is no way to disable it.

SELinux is a Linux Security Module (LSM) that provides a mechanism for mandatory access control (MAC). Processes that run as root with full capabilities are still subject to the mandatory policy restrictions.

The policy in Bottlerocket has the following objectives:

  1. Prevent most components from directly modifying the API settings.
  2. Block most components from modifying the container archives saved on disk.
  3. Stop containers from directly modifying the layers for other running containers.

The policy is currently aimed at hardening the OS against persistent threats. Future enhancements to the policy will focus on mitigating the impact of OS vulnerabilities, and protecting containers from other containers.

Kernel lockdown in integrity mode

Bottlerocket enables Lockdown in "integrity" mode by default on most variants.

Lockdown is a Linux Security Module (LSM) that blocks certain actions which could compromise the Linux kernel. As with SELinux, even processes that run as root with full capabilities are subject to these restrictions.

Certain variants such as *-nvidia need to load unsigned kernel modules at runtime. This is prohibited by the "integrity" mode, but required for the hardware to work as expected. On these variants, Lockdown is set to "none" instead.

Secure Boot enabled

Bottlerocket enables Secure Boot for all new variants on platforms that support UEFI boot.

The goal is to prevent unsigned, untrusted code from running at any point until containers are started. This is achieved by establishing the following chain of trust:

  1. The trusted platform firmware verifies that shim is signed correctly, then loads it.
  2. shim verifies that grub is signed correctly, then loads it.
  3. grub verifies that its grub.cfg is signed correctly, then loads it.
  4. grub verifies that the Linux kernel is signed correctly, then loads it.
  5. The Linux kernel verifies that the immutable root filesystem has not been altered.

Secure Boot only applies to platforms using UEFI firmware, and it is only enforced when the feature is enabled in the firmware. Therefore, systems using the legacy BIOS boot mode cannot benefit from Secure Boot. This includes Xen-based EC2 instance types, and bare metal machines configured to emulate the legacy BIOS boot mode.