This topic describes how you can configure Supply Chain Security Tools (SCST) - Policy. SCST - Policy requires extra configuration steps to verify your container images.
An image is admitted after it is validated against a policy with a matching image pattern and where at least one valid signature is obtained from the authorities provided in a matched ClusterImagePolicy
. For more information, see Create a ClusterImagePolicy resource later in the topic.
If more than one policy exists with a matching image pattern, all of the policies must have at least one passing authority for the image.
The Policy Controller only validates resources in include namespaces. Include a namespace resource by adding the label policy.sigstore.dev/include: "true"
. For example, by running:
$ kubectl label namespace my-secure-namespace policy.sigstore.dev/include=true
CautionWithout a Policy Controller
ClusterImagePolicy
applied, there are fallback behaviors where images are validated against the public Sigstore Rekor and Fulcio servers by using a keyless authority flow.Therefore, if the deploying image is signed publicly by a third party using the keyless authority flow, the image is admitted because it can validate against the public Rekor and Fulcio. To avoid this behavior, develop and apply a
ClusterImagePolicy
that applies to the images being deployed in the namespace.
The cluster image policy is a custom resource containing properties described in the following sections.
In a ClusterImagePolicy
, spec.images
specifies a list of glob matching patterns. These patterns are matched against the image digest in PodSpec
for resources attempting deployment.
Policy Controller defines the following globs by default:
*
is specified, the glob
matching behavior is index.docker.io/library/*
.*/*
is specified, the glob
matching behavior is index.docker.io/*/*
.With these default values, you require the glob
pattern **
to match against all images. If your image is hosted on Docker Hub, include index.docker.io
as the host for the glob.
This is a sample ClusterImagePolicy
that matches against all images using glob:
apiVersion: policy.sigstore.dev/v1beta1
kind: ClusterImagePolicy
metadata:
name: image-policy
spec:
images:
- glob: "**"
In a ClusterImagePolicy
, spec.mode
specifies the action of a policy:
enforce
: This is the default behavior. If the policy fails to validate the image, the policy fails.warn
: If the policy fails to validate the image, validation error messages are converted to warnings and the policy passes.This is a sample of a ClusterImagePolicy
that has warn
mode configured:
---
apiVersion: policy.sigstore.dev/v1beta1
kind: ClusterImagePolicy
metadata:
name: POLICY-NAME
spec:
mode: warn
Where POLICY-NAME
is the name of the policy you want to configure your ClusterImagePolicy
with.
When enforce
mode is set, an image that fails validation is not admitted.
Sample output message:
error: failed to patch: admission webhook "policy.sigstore.dev" denied the request: validation failed: failed policy: POLICY-NAME: spec.template.spec.containers[0].image
IMAGE-REFERENCE signature key validation failed for authority authority-0 for IMAGE-REFERENCE: GET IMAGE-SIGNATURE-REFERENCE: DENIED: denied; denied
failed policy: POLICY-NAME: spec.template.spec.containers[1].image
IMAGE-REFERENCE signature key validation failed for authority authority-0 for IMAGE-REFERENCE: GET IMAGE-SIGNATURE-REFERENCE: DENIED: denied; denied
When warn
mode is set, an image that fails validation is admitted.
Sample output message:
Warning: failed policy: POLICY-NAME: spec.template.spec.containers[0].image
Warning: IMAGE-REFERENCE signature key validation failed for authority authority-0 for IMAGE-REFERENCE: GET IMAGE-SIGNATURE-REFERENCE: DENIED: denied; denied
Warning: failed policy: POLICY-NAME: spec.template.spec.containers[1].image
Warning: IMAGE-REFERENCE signature key validation failed for authority authority-0 for IMAGE-REFERENCE: GET IMAGE-SIGNATURE-REFERENCE: DENIED: denied; denied
If you do not want a Warning
output message, you can configure a static.action
pass
authority to allow expected unsigned images. For example, you might want to allow unsigned images if your policy controller runs on a development environment and you need to iterate quickly. For information about static action authorities, see Static Action.
You can use match
to filter resources by group, version, kind, or labels in a selected namespace to enforce the defined policy. If the list of matching resources is empty, all core resources are used by default.
For example, you can filter all v1
cronjobs
with the label app: tap
in a namespace that is labeled for policy enforcement:
spec:
match:
- group: batch
resource: cronjobs
version: v1
selector:
matchLabels:
app: tap
Authorities listed in the authorities
block of the ClusterImagePolicy
are key
or keyless
specifications.
Each key
authority can contain a PEM-encoded ECDSA public key, a secretRef
, or a kms
path. For example:
spec:
authorities:
- key:
data: |
-----BEGIN PUBLIC KEY-----
...
-----END PUBLIC KEY-----
- key:
secretRef:
name: secretName
- key:
kms: KMSPATH
Where KMSPATH
is the name of the KMS path you want to configure in your key authority.
ImportantOnly ECDSA public keys are supported.
The policy resyncs with KMS referenced every 10 hours. Any updates to the secret in KMS is pulled in during the refresh. To force a resync, the policy must be deleted and re-created.
The secret referenced in key.secretRef.name
must be created in the cosign-system
namespace or the namespace where the Policy Controller is installed. This secret must only contain one data
entry with the public key.
Each keyless authority can contain a Fulcio URL, a Rekor URL, a certificate, or an array of identities.
NoteKeyless support is deactivated by default. For more information, see Install Supply Chain Security Tools - Policy Controller.
Identities are represented with a combination of issuer
or issuerRegExp
with subject
or subjectRegExp
.
issuer
: Defines the issuer for this identity.issuerRegExp
: Specifies a regular expression to match the issuer for this identity.subject
: Defines the subject for this identity.subjectRegExp
: Specifies a regular expression to match the subject for this identity.An example of keyless authority structure:
spec:
authorities:
- keyless:
url: https://fulcio.example.com
ca-cert:
data: Certificate Data
identities:
- issuer: https://accounts.google.com
subjectRegExp: .*@example.com
- issuer: https://token.actions.githubusercontent.com
subject: https://github.com/mycompany/*/.github/workflows/*@*
ctlog:
url: https://rekor.example.com
- keyless:
url: https://fulcio.example.com
ca-cert:
secretRef:
name: secretName
identities:
- issuerRegExp: .*kubernetes.default.*
subjectRegExp: .*kubernetes.io/namespaces/default/serviceaccounts/default
The authorities are evaluated by using the any of
operator to admit container images. For each pod, the Policy Controller iterates over the list of containers and init containers. For every policy that matches against the images, they must each have at least one valid signature obtained by using the authorities specified. If an image does not match any policy, the Policy Controller does not admit the image.
ClusterImagePolicy
authorities are configured to always pass
or fail
with static.action
.
This is a sample ClusterImagePolicy
with static action fail
:
apiVersion: policy.sigstore.dev/v1beta1
kind: ClusterImagePolicy
metadata:
name: POLICY-NAME
spec:
authorities:
- static:
action: fail
Where POLICY-NAME
is the name of the policy you want to configure your ClusterImagePolicy
with.
This is sample output of static action fail
:
error: failed to patch: admission webhook "policy.sigstore.dev" denied the request: validation failed: failed policy: POLICY-NAME: spec.template.spec.containers[0].image
IMAGE-REFERENCE disallowed by static policy
failed policy: POLICY-NAME: spec.template.spec.containers[1].image
IMAGE-REFERENCE disallowed by static policy
Images that are unsigned in a namespace with validation enabled are admitted with an authority with the static action pass
.
This applies when you are configuring a policy with static.action
pass
for tap-packages
images. Another policy is then configured to validate signed images produced by Tanzu Build Service. This allows images from tap-packages
, which are unsigned and required by the platform, to be admitted while still validating signed built images from Tanzu Build Service. For more information, see Configure your supply chain to sign and verify your image builds.
If you want Warning
messages for admitted images where validation failed, you can configure a policy with warn
mode and valid authorities. For information about ClusterImagePolicy
modes, see Mode earlier in this topic.
There are three ways the package reads credentials to authenticate to registries protected by authentication:
Reading imagePullSecrets
directly from the resource being admitted. For more information, see Container image pull secrets in the Kubernetes documentation.
Reading imagePullSecrets
from the service account the resource is running as. For more information, see Arranging for imagePullSecrets to be automatically attached in the Kubernetes documentation.
Reading a secretRef
from the ClusterImagePolicy
resource’s signaturePullSecrets
when specifying the cosign signature source.
Authentication can fail for the following scenarios:
imagePullSecrets
of the resource or in the service account the resource runs as.ClusterImagePolicy
signaturePullSecrets
field.You can provide secrets for authentication as part of the policy configuration. The oci
location is the image location or a remote location where signatures are configured to be stored during signing. signaturePullSecrets
is available in the cosign-system
namespace or the namespace where the Policy Controller is installed.
By default, imagePullSecrets
from the resource or service account is used while the default oci
location is the image location.
See the following example:
spec:
authorities:
- key:
data: |
-----BEGIN PUBLIC KEY-----
...
-----END PUBLIC KEY-----
source:
- oci: registry.example.com/project/signature-location
signaturePullSecrets:
- name: MY-SECRET
- keyless:
url: https://fulcio.example.com
source:
- oci: registry.example.com/project/signature-location
signaturePullSecrets:
- name: MY-SECRET
Where MY-SECRET
is the name of the secret you want to use with your credentials.
VMware recommends using a set of credentials with the least amount of privilege that allows reading the signature stored in your registry.
A sample policy:
apiVersion: policy.sigstore.dev/v1beta1
kind: ClusterImagePolicy
metadata:
name: image-policy
spec:
images:
- glob: "gcr.io/projectsigstore/cosign*"
authorities:
- name: official-cosign-key
key:
data: |
-----BEGIN PUBLIC KEY-----
MFkwEwYHKoZIzj0CAQYIKoZIzj0DAQcDQgAEhyQCx0E9wQWSFI9ULGwy3BuRklnt
IqozONbbdbqz11hlRJy9c7SG+hdcFl9jE9uE/dwtuwU2MqU9T/cN0YkWww==
-----END PUBLIC KEY-----
When using the sample policy, run these commands to verify your configuration:
Verify that the Policy Controller admits the signed image that validates with the configured public key by running:
kubectl run cosign \
--image=gcr.io/projectsigstore/cosign:v1.2.1 \
--dry-run=server
For example:
$ kubectl run cosign \
--image=gcr.io/projectsigstore/cosign:v1.2.1 \
--dry-run=server
pod/cosign created (server dry run)
If you are using vSphere IaaS control plane (formerly named vSphere with Tanzu) or OpenShift, you must add some overrides as in this example:
$ kubectl run cosign \
--image=gcr.io/projectsigstore/cosign:v1.2.1 \
--overrides='{"spec": {"securityContext": {"seccompProfile": {"type": "RuntimeDefault"}}, "containers": [{"name": "cosign", "securityContext": {"allowPrivilegeEscalation": false, "runAsNonRoot": true, "capabilities": {"drop": ["ALL"]}}}]}}' \
--override-type strategic \
--dry-run=server
pod/cosign created (server dry run)
Verify that the Policy Controller rejects the unmatched image by running:
kubectl run busybox --image=busybox --dry-run=server
For example:
$ kubectl run busybox --image=busybox --dry-run=server
Error from server (BadRequest): admission webhook "policy.sigstore.dev" denied the request: validation failed: no matching policies: spec.containers[0].image
index.docker.io/library/busybox@sha256:3614ca5eacf0a3a1bcc361c939202a974b4902b9334ff36eb29ffe9011aaad83
In the example output, it did not specify which authorities were used as there was no policy found that matched the image. Therefore the image fails to validate for a signature and fails to deploy.
Verify that the Policy Controller rejects a matched image signed with a different key than the one configured by running:
kubectl run cosign-fail \
--image=gcr.io/projectsigstore/cosign:v0.3.0 \
--dry-run=server
For example:
$ kubectl run cosign-fail \
--image=gcr.io/projectsigstore/cosign:v0.3.0 \
--dry-run=server
Error from server (BadRequest): admission webhook "policy.sigstore.dev" denied the request: validation failed: failed policy: image-policy: spec.containers[0].image
gcr.io/projectsigstore/cosign@sha256:135d8c5e27bdc917f04b415fc947d7d5b1137f99bb8fa00bffc3eca1856e9c52 failed to validate public keys with authority official-cosign-key for gcr.io/projectsigstore/cosign@sha256:135d8c5e27bdc917f04b415fc947d7d5b1137f99bb8fa00bffc3eca1856e9c52: no matching signatures:
In the example output, it specifies which authorities were used for validation when a policy was found that matched the image. In this case, the authority used was official-cosign-key
. If no name is specified, it is authority-#
by default.