kubeaudit
is a command line tool to audit Kubernetes clusters for various
different security concerns: run the container as a non-root user, use a read
only root filesystem, drop scary capabilities, don't add new ones, don't run
privileged, ... You get the gist of it and more on that later. Just know:
Kubeaudit has official releases that are blessed and stable here: Official releases
Master will have newer features than the stable releases. If you need a newer feature not yet included in a release you can do the following to get kubeaudit:
go get -v github.com/Shopify/kubeaudit
make
make install
Now you can just call kubeaudit
with one of commands from here
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
just by
- running
make plugin
and having $GOPATH/bin available in your path.
or
- renaming the binary to
kubectl-audit
and having it available in your path.
kubeaudit
has three different modes for its audits:
- Cluster mode
If kubeaudit detects that it's running in a container,
kubeaudit cmd
will attempt to audit the cluster it's running in. - Local config mode
If kubeaudit is not running in a container,
kubeaudit cmd
will audit the resources specified by your local kubeconfig ($HOME/.kube/config
) file. You can force kubeaudit to use a specific local config file with the switch-c/--kubeconfig /config/path
- Manifest mode
If you wish to audit a manifest file, use the command
kubeaudit -f/--manifest /path/to/manifest.yml
kubeaudit
supports two different output types:
- just running
kubeaudit
will log human readable output - if run with
-j/--json
it will log output json formatted so that its output can be used by other programs easily
kubeaudit
has four different log levels INFO, WARN, ERROR
controlled by
-v/--verbose LEVEL
and for those who counted and want to work on kubeaudit
DEBUG
- by default the debug level is set to
ERROR
and will logINFO
,WARN
andERROR
- if you only care about
ERROR
set it toERROR
- if you care about
ERROR
andWARN
set it toWARN
But wait! Which version am I actually running? kubeaudit version
will tell you.
I need help! Run kubeaudit help
every audit has its own help so you can run
kubeaudit help sc
Last but not least before we look at the audits: kubeaudit -a/--allPods
audits against pods in all the phases (default Running Phase)
As humans we are lazy and kubeaudit
knows that so it comes with the functionality to autofix
workload manifests. Point it at your workload manifests and it will automagically fix everything so that manifests are as secure as it gets.
kubeaudit autofix -f path/to/manifest.yml
The manifest might end up a little too secure for the work it is supposed to do. If that is the case check out labels to opt out of certain checks.
kubeaudit
has multiple checks:
- Audit all
- Audit security context
- Audit image
- Audit Service Accounts
- Audit network policies
- Audit resources
Runs all the above checks.
kubeaudit all
ERRO[0000] RunAsNonRoot is not set, which results in root user being allowed!
ERRO[0000] Default serviceAccount with token mounted. Please set automountServiceAccountToken to false
WARN[0000] Privileged defaults to false, which results in non privileged, which is okay.
ERRO[0000] Capability not dropped CapName=AUDIT_WRITE
The security context holds a couple of different security related
configurations. For convenience, kubeaudit
will always log the following
information when it creates a log:
kubeaudit command
LOG[0000] KubeType=deployment Name=THEdeployment Namespace=deploymentNS
And for brevity, the information will not be shown in the commands below.
Currently, kubeaudit
is able to check for the following fields in the security context:
kubeaudit
will detect whether readOnlyRootFilesystem
is either not set nil
or explicitly set to false
kubeaudit rootfs
ERRO[0000] ReadOnlyRootFilesystem not set which results in a writable rootFS, please set to true
ERRO[0000] ReadOnlyRootFilesystem set to false, please set to true
kubeaudit
will detect whether the container is to be run as root:
kubeaudit nonroot
ERRO[0000] RunAsNonRoot is set to false (root user allowed), please set to true!
ERRO[0000] RunAsNonRoot is not set, which results in root user being allowed!
kubeaudit
will detect whether allowPrivilegeEscalation
is either set to nil
or explicitly set to false
kubeaudit allowpe
ERRO[0000] AllowPrivilegeEscalation set to true, please set to false
ERRO[0000] AllowPrivilegeEscalation not set which allows privilege escalation, please set to false
kubeaudit
will detect whether the container is to be run privileged:
kubeaudit priv
ERRO[0000] Privileged set to true! Please change it to false!
Since we want to make sure everything is intentionally configured correctly kubeaudit
warns about privileged
not being set:
kubeaudit priv
WARN[0000] Privileged defaults to false, which results in non privileged, which is okay.
Docker comes with a couple of capabilities that shouldn't be needed and
therefore should be dropped. kubeaudit
will also complain about added capabilities.
If the capabilities field doesn't exist within the security context:
kubeaudit caps
ERRO[0000] Capabilities field not defined!
When capabilities were added:
kubeaudit caps
ERRO[0000] Capability added CapName=NET_ADMIN
config/caps
holds a list of capabilities that we recommend be dropped, change it if you
want to keep some of the capabilities otherwise kubeaudit
will complain about
them not being dropped:
kubeaudit caps
ERRO[0000] Capability not dropped CapName=AUDIT_WRITE
kubeaudit
can check for image names and image tags:
-
If the image tag is incorrect an ERROR will issued
kubeaudit image -i gcr.io/google_containers/echoserver:1.7 ERRO[0000] Image tag was incorrect
-
If the image doesn't have a tag but an image of the name was found a WARNING will be created:
kubeaudit image -i gcr.io/google_containers/echoserver:1.7 WARN[0000] Image tag was missing
-
If the image was found with correct tag
kubeaudit
notifies with an INFO message:kubeaudit image -i gcr.io/google_containers/echoserver:1.7 INFO[0000] Image tag was correct
It audits against the following scenarios:
-
A default serviceAccount mounted with a token:
kubeaudit sat ERRO[0000] Default serviceAccount with token mounted. Please set AutomountServiceAccountToken to false
-
A deprecated service account:
kubeaudit sat WARN[0000] serviceAccount is a deprecated alias for ServiceAccountName, use that one instead DSA=DeprecatedServiceAccount
It checks that every namespace should have a default deny network policy installed. See Kubernetes Network Policies for more information:
kubeaudit np
WARN[0000] Default allow mode on test/testing
It checks that every resource has a CPU and memory limit. See Kubernetes Resource Quotas for more information:
kubeaudit limits
WARN[0000] CPU limit not set, please set it!
WARN[0000] Memory limit not set, please set it!
With the --cpu
and --memory
parameters, kubeaudit
can check the limits not to be exceeded.
kubeaudit limits --cpu 500m --memory 125Mi
WARN[0000] CPU limit exceeded, it is set to 1 but it must not exceed 500m. Please adjust it! !
WARN[0000] Memory limit exceeded, it is set to 512Mi but it must not exceed 125Mi. Please adjust it!
It checks that AppArmor is enabled for all containers by making sure the following annotation exists on the pod. There must be an annotation for each container in the pod:
container.apparmor.security.beta.kubernetes.io/<container name>: <profile>
where profile can be "runtime/default" or start with "localhost/" to be considered valid.
If the AppArmor annotation is missing:
kubeaudit apparmor
ERRO[0000] AppArmor annotation missing. Container=myContainer KubeType=pod Name=myPod Namespace=myNamespace
When AppArmor annotations are misconfigured:
kubeaudit apparmor
ERRO[0000] AppArmor disabled. Annotation=container.apparmor.security.beta.kubernetes.io/myContainer
Container=myContainer KubeType=pod Name=myPod Namespace=myNamespace Reason=badval
It checks that Seccomp is enabled for all containers by making sure one or both of the following annotations exists on the pod. If no pod annotation is used, then there must be an annotation for each container. Container annotations override the pod annotation:
# pod annotation
seccomp.security.alpha.kubernetes.io/pod: <profile>
# container annotation
container.seccomp.security.alpha.kubernetes.io/<container name>: <profile>
where profile can be "runtime/default" or start with "localhost/" to be considered valid. "docker/default" is deprecated and will show a warning. It should be replaced with "runtime/default".
If the Seccomp annotation is missing:
kubeaudit seccomp
ERRO[0000] Seccomp annotation missing. Container=myContainer KubeType=pod Name=myPod Namespace=myNamespace
When Seccomp annotations are misconfigured for a container:
kubeaudit seccomp
ERRO[0000] Seccomp disabled for container. Annotation=container.seccomp.security.alpha.kubernetes.io/myContainer
Container=myContainer KubeType=pod Name=myPod Namespace=myNamespace Reason=badval
When Seccomp annotations are misconfigured for a pod:
kubeaudit seccomp
ERRO[0000] Seccomp disabled for pod. Annotation=seccomp.security.alpha.kubernetes.io/pod Container= KubeType=pod
Name=myPod Namespace=myNamespace Reason=unconfined
Override labels give you the ability to have kubeaudit
allow certain audits to fail.
For example, if you want kubeaudit
to ignore the fact that AllowPrivilegeEscalation
was set to true
, you can add the following label:
spec:
template:
metadata:
labels:
apps: YourAppNameHere
audit.kubernetes.io/allow-privilege-escalation: "YourReasonForOverrideHere"
Any label with a non-nil reason string will prevent kubeaudit
from throwing the corresponding error and issue a warning instead.
Reasons matching "true"
(not case sensitive) will be displayed as Unspecified
.
kubeaudit
supports many labels:
- audit.kubernetes.io/allow-privilege-escalation
- audit.kubernetes.io/allow-privileged
- audit.kubernetes.io/allow-capability
- audit.kubernetes.io/allow-run-as-root
- audit.kubernetes.io/allow-automount-service-account-token
- audit.kubernetes.io/allow-read-only-root-filesystem-false
Allows allowPrivilegeEscalation
to be set to true
.
kubeaudit.allow.privilegeEscalation: "Superuser privileges needed"
WARN[0000] Allowed setting AllowPrivilegeEscalation to true Reason="Superuser privileges needed"
Allows privileged
to be set to true
.
kubeaudit.allow.privileged: "Privileged execution required"
WARN[0000] Allowed setting privileged to true Reason="Privileged execution required"
Allows adding a capability or keeping one that would otherwise be dropped.
kubeaudit.allow.capability.chown: "true"
WARN[0000] Capability allowed CapName=CHOWN Reason=Unspecified
Allows setting runAsNonRoot
to false
.
kubeaudit.allow.runAsRoot: "Root privileges needed"
WARN[0000] Allowed setting RunAsNonRoot to false Reason="Root privileges needed"
Allows setting automountServiceAccountToken
to true
.
kubeaudit.allow.autmountServiceAccountToken: "True"
WARN[0000] Allowed setting automountServiceAccountToken to true Reason=Unspecified
Allows setting readOnlyRootFilesystem
to false
kubeaudit.allow.readOnlyRootFilesystemFalse: "Write permissions needed"
WARN[0000] Allowed setting readOnlyRootFilesystem to false Reason="Write permissions needed"
Allows configuring the audit against drop capabilities. Sane defaults are as follows:
# SANE DEFAULTS:
capabilitiesToBeDropped:
# https://docs.docker.com/engine/reference/run/#runtime-privilege-and-linux-capabilities
- SETPCAP #Modify process capabilities.
- MKNOD #Create special files using mknod(2).
- AUDIT_WRITE #Write records to kernel auditing log.
- CHOWN #Make arbitrary changes to file UIDs and GIDs (see chown(2)).
- NET_RAW #Use RAW and PACKET sockets.
- DAC_OVERRIDE #Bypass file read, write, and execute permission checks.
- FOWNER #Bypass permission checks on operations that normally require the file system UID of the process to match the UID of the file.
- FSETID #Don’t clear set-user-ID and set-group-ID permission bits when a file is modified.
- KILL #Bypass permission checks for sending signals.
- SETGID #Make arbitrary manipulations of process GIDs and supplementary GID list.
- SETUID #Make arbitrary manipulations of process UIDs.
- NET_BIND_SERVICE #Bind a socket to internet domain privileged ports (port numbers less than 1024).
- SYS_CHROOT #Use chroot(2), change root directory.
- SETFCAP #Set file capabilities.
This can be overridden by using -d
flag and providing your own defaults in the yaml format as shown above.
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
- Run the tests to see everything is working as expected:
make test
- Commit your changes:
git commit -am 'Adds awesome feature'
- Push to the branch:
git push fork
- Submit a PR
- ???
- Profit