Custom Checks with Rego
Trivy can scan configuration files for common security issues (a.k.a IaC misconfiguration scanning). In addition to a comprehensive built in database of checks, you can add your own custom checks. Checks are written in Rego language and the full documentation for checks and customizing them is available here.
This tutorial will walk you through writing a custom check in Rego that checks for an issue in a Dockerfile.
When you are writing a check, it's important to understand the input to the check. This will be the IaC file that you are scanning; for example, a Kubernetes YAML resource definition, or an AWS JSON CloudFormation, or in our case a Dockerfile.
Since Rego is primarily tailored to query JSON objects, all incoming configuration files needs to be first converted to structured objects, which is available to the Rego code as the input variable. This is nothing that users have to do manually in Trivy. Instead, Rego makes it possible to pass in custom Schemas that detail how files are converted. Once Rego has access to a custom Schema, it will know in which format to access configuration files such as a Dockerfile.
Here you can find the schemas that define how different configuration files are converted to JSON by Trivy. This tutorial will make use of the dockerfile.json schema. The schema will need to be parsed into your custom check.
Users can also use the Schema Explorer to view the structure of the data provided to Rego.
Create a Rego file and Specify Trivy metadata
First, create a new .rego
file e.g. a docker-check.rego
file:
touch docker-check.rego
Next, we need to specify metadata about the check. This is information that helps Trivy load and process the check.
# METADATA
# title: Verify Image
# description: Verify Image is allowed to be used and in the right format
# schemas:
# - input: schema["dockerfile"]
# custom:
# id: ID001
# severity: MEDIUM
# input:
# selector:
# - type: dockerfile
Important: The METADATA
has to be defined on top of the file.
More information on the different fields in the metadata can be found in the Trivy documentation.
Package and imports
package custom.dockerfile.ID001
import future.keywords.in
Every rego check has a package name. In our case, we will call it custom.dockerfile.ID001
to avoid confusion between custom checks and built-in checks. The group name dockerfile
has no effect on the package name. Note that each package has to contain only one check. However, we can pass multiple checks into our Trivy scan.
The first keyword of the package, in this case custom
, will be reused in the trivy
command as the --namespace
.
Allowed data
The check that we are setting up compares the container images used in the Dockerfile with a list of white-listed container images. Thus, we need to add the images that are allowed to be used in the Dockerfile to our check. In our case, we will store them in an array of arrays:
allowed_images := {
["node:21-alpine3.19", "as", "build-deps"],
["nginx:1.2"]
}
Select the images that are used in the Dockerfile
Next, we need to iterate over the different commands in our Dockerfile and identify the commands that provide the base container images:
deny[msg] {
input.Stages[m].Commands[l].Cmd == "from"
val := input.Stages[m].Commands[l].Value
not val in allowed_images
msg := sprintf("The container image '%s' used in the Dockerfile is not allowed", val)
}
Let's look at the check line by line:
- The rule should always be
deny
in the Trivy Rego checks input.Stages[m].Commands[l].Cmd
input
allows us to access the different commands in the Dockerfile. We need to access the commands that use "FROM". Every command will be converted to lowercase.val := input.Stages[m].Commands[l].Value
accesses the value of theFROM
command and stores it inval
not val in allowed_images
checks whether val is not part of our allowed images list; this part of the check relies on the import statement- In case our check fails, the
msg
will be printed with the image name used inval
Note that Rego
- uses
AND
automatically to combine conditions in this check - automatically iterates through the array of commands in the Dockefile and allowed images
Run the check in a Trivy misconfiguration scan
Ensure that you have Trivy installed and run the following command:
trivy fs --scanners misconf --config-check ./docker-check.rego --namespaces custom ./Dockerfile
Please replace:
./docker-check.rego
with the file path to your checkcustom
should be replaced with your package name if different./Dockerfile
is the path to the Dockerfile that should be scanned
Note: If you define custom packages, you have to specify the package prefix via --namespaces
option. In our case, we called the custom package custom
.
Resources
- Rego provides a long list of courses that can be useful in writing more complex checks
- The Rego documentation provides detailed information on the different types, iterations etc.
- Have a look at the built-in checks for Trivy for inspiration on how to write custom checks.