kubeaudit
is a command line tool and a Go package to audit Kubernetes clusters for various different security concerns, such as:
- run as non-root
- use a read-only root filesystem
- drop scary capabilities, don't add new ones
- don't run privileged
- and more!
tldr. kubeaudit
makes sure you deploy secure containers!
Package
To use kubeaudit as a Go package, see the package docs.
The rest of this README will focus on how to use kubeaudit as a command line tool.
Command Line Interface (CLI)
Installation
Brew
Download a binary
Kubeaudit has official releases that are blessed and stable: Official releases
DIY build
Master may have newer features than the stable releases. If you need a newer feature not yet included in a release, make sure you're using Go 1.17+ and run the following:
go get -v github.com/Shopify/kubeaudit
Start using kubeaudit
with the Quick Start or view all the supported commands.
Kubectl Plugin
Prerequisite: kubectl v1.12.0 or later
With kubectl v1.12.0 introducing easy pluggability of external functions, kubeaudit can be invoked as kubectl audit
by
- running
make plugin
and having$GOPATH/bin
available in your path.
or
- renaming the binary to
kubectl-audit
and having it available in your path.
Docker
We also release a Docker image: shopify/kubeaudit
. To run kubeaudit as a job in your cluster see Running kubeaudit in a cluster.
Quick Start
kubeaudit has three modes:
- Manifest mode
- Local mode
- Cluster mode
Manifest Mode
If a Kubernetes manifest file is provided using the -f/--manifest
flag, kubeaudit will audit the manifest file.
Example command:
kubeaudit all -f "/path/to/manifest.yml"
Example output:
$ kubeaudit all -f "internal/test/fixtures/all_resources/deployment-apps-v1.yml"---------------- Results for ---------------
apiVersion: apps/v1
kind: Deployment
metadata:
name: deployment
namespace: deployment-apps-v1
--------------------------------------------
-- [error] AppArmorAnnotationMissing
Message: AppArmor annotation missing. The annotation 'container.apparmor.security.beta.kubernetes.io/container' should be added.
Metadata:
Container: container
MissingAnnotation: container.apparmor.security.beta.kubernetes.io/container
-- [error] AutomountServiceAccountTokenTrueAndDefaultSA
Message: Default service account with token mounted. automountServiceAccountToken should be set to 'false' or a non-default service account should be used.
-- [error] CapabilityShouldDropAll
Message: Capability not set to ALL. Ideally, you should drop ALL capabilities and add the specific ones you need to the add list.
Metadata:
Container: container
Capability: AUDIT_WRITE
...
If no errors with a given minimum severity are found, the following is returned:
Autofix
Manifest mode also supports autofixing all security issues using the autofix
command:
kubeaudit autofix -f "/path/to/manifest.yml"
To write the fixed manifest to a new file instead of modifying the source file, use the -o/--output
flag.
kubeaudit autofix -f "/path/to/manifest.yml" -o "/path/to/fixed"
To fix a manifest based on custom rules specified on a kubeaudit config file, use the -k/--kconfig
flag.
kubeaudit autofix -k "/path/to/kubeaudit-config.yml" -f "/path/to/manifest.yml" -o "/path/to/fixed"
Cluster Mode
Kubeaudit can detect if it is running within a container in a cluster. If so, it will try to audit all Kubernetes resources in that cluster:
Local Mode
Kubeaudit will try to connect to a cluster using the local kubeconfig file ($HOME/.kube/config
). A different kubeconfig location can be specified using the --kubeconfig
flag. To specify a context of the kubeconfig, use the -c/--context
flag.
kubeaudit all --kubeconfig "/path/to/config" --context my_cluster
For more information on kubernetes config files, see https://kubernetes.io/docs/concepts/configuration/organize-cluster-access-kubeconfig/
Audit Results
Kubeaudit produces results with three levels of severity:
Error
: A security issue or invalid kubernetes configurationWarning
: A best practice recommendationInfo
: Informational, no action required. This includes results that are overridden
The minimum severity level can be set using the --minSeverity/-m
flag.
By default kubeaudit will output results in a human-readable way. If the output is intended to be further processed, it can be set to output JSON using the --format json
flag. To output results as logs (the previous default) use --format logrus
. Some output formats include colors to make results easier to read in a terminal. To disable colors (for example, if you are sending output to a text file), you can use the --no-color
flag.
If there are results of severity level error
, kubeaudit will exit with exit code 2. This can be changed using the --exitcode/-e
flag.
For all the ways kubeaudit can be customized, see Global Flags.
Commands
Command | Description | Documentation |
---|---|---|
all | Runs all available auditors, or those specified using a kubeaudit config. | docs |
autofix | Automatically fixes security issues. | docs |
version | Prints the current kubeaudit version. |
Auditors
Auditors can also be run individually.
Command | Description | Documentation |
---|---|---|
apparmor | Finds containers running without AppArmor. | docs |
asat | Finds pods using an automatically mounted default service account | docs |
capabilities | Finds containers that do not drop the recommended capabilities or add new ones. | docs |
deprecatedapis | Finds any resource defined with a deprecated API version. | docs |
hostns | Finds containers that have HostPID, HostIPC or HostNetwork enabled. | docs |
image | Finds containers which do not use the desired version of an image (via the tag) or use an image without a tag. | docs |
limits | Finds containers which exceed the specified CPU and memory limits or do not specify any. | docs |
mounts | Finds containers that have sensitive host paths mounted. | docs |
netpols | Finds namespaces that do not have a default-deny network policy. | docs |
nonroot | Finds containers running as root. | docs |
privesc | Finds containers that allow privilege escalation. | docs |
privileged | Finds containers running as privileged. | docs |
rootfs | Finds containers which do not have a read-only filesystem. | docs |
seccomp | Finds containers running without Seccomp. | docs |
Global Flags
Short | Long | Description |
---|---|---|
--format | The output format to use (one of "pretty", "logrus", "json") (default is "pretty") | |
--kubeconfig | Path to local Kubernetes config file. Only used in local mode (default is $HOME/.kube/config ) | |
-c | --context | The name of the kubeconfig context to use |
-f | --manifest | Path to the yaml configuration to audit. Only used in manifest mode. You may use - to read from stdin. |
-n | --namespace | Only audit resources in the specified namespace. Not currently supported in manifest mode. |
-g | --includegenerated | Include generated resources in scan (such as Pods generated by deployments). If you would like kubeaudit to produce results for generated resources (for example if you have custom resources or want to catch orphaned resources where the owner resource no longer exists) you can use this flag. |
-m | --minseverity | Set the lowest severity level to report (one of "error", "warning", "info") (default is "info") |
-e | --exitcode | Exit code to use if there are results with severity of "error". Conventionally, 0 is used for success and all non-zero codes for an error. (default is 2) |
--no-color | Don't use colors in the output (default is false) |
Configuration File
The kubeaudit config can be used for two things:
- Enabling only some auditors
- Specifying configuration for auditors
Any configuration that can be specified using flags for the individual auditors can be represented using the config.
The config has the following format:
enabledAuditors:
# Auditors are enabled by default if they are not explicitly set to "false"
apparmor: false
asat: false
capabilities: true
deprecatedapis: true
hostns: true
image: true
limits: true
mounts: true
netpols: true
nonroot: true
privesc: true
privileged: true
rootfs: true
seccomp: true
auditors:
capabilities:
# add capabilities needed to the add list, so kubeaudit won't report errors
allowAddList: ['AUDIT_WRITE', 'CHOWN']
deprecatedapis:
# If no versions are specified and the'deprecatedapis' auditor is enabled, WARN
# results will be genereted for the resources defined with a deprecated API.
currentVersion: '1.22'
targetedVersion: '1.25'
image:
# If no image is specified and the 'image' auditor is enabled, WARN results
# will be generated for containers which use an ima ge without a tag
image: 'myimage:mytag'
limits:
# If no limits are specified and the 'limits' auditor is enabled, WARN results
# will be generated for containers which have no cpu or memory limits specified
cpu: '750m'
memory: '500m'
For more details about each auditor, including a description of the auditor-specific configuration in the config, see the Auditor Docs.
Note: The kubeaudit config is not the same as the kubeconfig file specified with the --kubeconfig
flag, which refers to the Kubernetes config file (see Local Mode). Also note that only the all
and autofix
commands support using a kubeaudit config. It will not work with other commands.
Note: If flags are used in combination with the config file, flags will take precedence.
Override Errors
Security issues can be ignored for specific containers or pods by adding override labels. This means the auditor will produce info
results instead of error
results and the audit result name will have Allowed
appended to it. The labels are documented in each auditor's documentation, but the general format for auditors that support overrides is as follows:
An override label consists of a key
and a value
.
The key
is a combination of the override type (container or pod) and an override identifier
which is unique to each auditor (see the docs for the specific auditor). The key
can take one of two forms depending on the override type:
- Container overrides, which override the auditor for that specific container, are formatted as follows:
container.audit.kubernetes.io/[container name].[override identifier]
- Pod overrides, which override the auditor for all containers within the pod, are formatted as follows:
audit.kubernetes.io/pod.[override identifier]
If the value
is set to a non-empty string, it will be displayed in the info
result as the OverrideReason
:
$ kubeaudit asat -f "auditors/asat/fixtures/service-account-token-true-allowed.yml"---------------- Results for ---------------
apiVersion: v1
kind: ReplicationController
metadata:
name: replicationcontroller
namespace: service-account-token-true-allowed
--------------------------------------------
-- [info] AutomountServiceAccountTokenTrueAndDefaultSAAllowed
Message: Audit result overridden: Default service account with token mounted. automountServiceAccountToken should be set to 'false' or a non-default service account should be used.
Metadata:
OverrideReason: SomeReason
As per Kubernetes spec, value
must be 63 characters or less and must be empty or begin and end with an alphanumeric character ([a-z0-9A-Z]
) with dashes (-
), underscores (_
), dots (.
), and alphanumerics between.
Multiple override labels (for multiple auditors) can be added to the same resource.
See the specific auditor docs for the auditor you wish to override for examples.
To learn more about labels, see https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
Contributing
If you'd like to fix a bug, contribute a feature or just correct a typo, please feel free to do so as long as you follow our Code of Conduct.
- Create your own fork!
- Get the source:
go get github.com/Shopify/kubeaudit
- Go to the source:
cd $GOPATH/src/github.com/Shopify/kubeaudit
- Add your forked repo as a fork:
git remote add fork https://github.com/you-are-awesome/kubeaudit
- Create your feature branch:
git checkout -b awesome-new-feature
- Install Kind
- Run the tests to see everything is working as expected:
make test
(to run tests without Kind:USE_KIND=false make test
) - Commit your changes:
git commit -am 'Adds awesome feature'
- Push to the branch:
git push fork
- Sign the Contributor License Agreement
- Submit a PR (All PR must be labeled with
(Bug fix),(New feature),
(Documentation update), or
(Breaking changes) )
- ???
- Profit
Note that if you didn't sign the CLA before opening your PR, you can re-run the check by adding a comment to the PR that says "I've signed the CLA!"!